Management evaluation of metastasis in the brain (MEMBRAIN)—a United Kingdom and Ireland prospective, multicenter observational study

Abstract

Background

In recent years an increasing number of patients with cerebral metastasis (CM) have been referred to the neuro-oncology multidisciplinary team (NMDT). Our aim was to obtain a national picture of CM referrals to assess referral volume and quality and factors affecting NMDT decision making.

Methods

A prospective multicenter cohort study including all adult patients referred to NMDT with 1 or more CM was conducted. Data were collected in neurosurgical units from November 2017 to February 2018. Demographics, primary disease, KPS, imaging, and treatment recommendation were entered into an online database.

Results

A total of 1048 patients were analyzed from 24 neurosurgical units. Median age was 65 years (range, 21-93 years) with a median number of 3 referrals (range, 1-17 referrals) per NMDT. The most common primary malignancies were lung (36.5%, n = 383), breast (18.4%, n = 193), and melanoma (12.0%, n = 126). A total of 51.6% (n = 541) of the referrals were for a solitary metastasis and resulted in specialist intervention being offered in 67.5% (n = 365) of cases. A total of 38.2% (n = 186) of patients being referred with multiple CMs were offered specialist treatment. NMDT decision making was associated with number of CMs, age, KPS, primary disease status, and extent of extracranial disease (univariate logistic regression, P < .001) as well as sentinel location and tumor histology (P < .05). A delay in reaching an NMDT decision was identified in 18.6% (n = 195) of cases.

Conclusions

This study demonstrates a changing landscape of metastasis management in the United Kingdom and Ireland, including a trend away from adjuvant whole-brain radiotherapy and specialist intervention being offered to a significant proportion of patients with multiple CMs. Poor quality or incomplete referrals cause delay in NMDT decision making.

The National Institute of Health and Care Excellence (NICE)¹ Improving Outcomes Guidance for brain and CNS tumors of 2006 recommended that management of all patients with brain tumors should be guided by a neuro-oncology multidisciplinary team (NMDT) to ensure consensus opinion on patient care is reached.² Because cerebral metastasis (CM) referrals to the weekly NMDT originate from a variety of sources, including the local emergency department, district general hospital, oncologists, or general practitioners, and NMDT members have not seen these patients a priori, the provided referral information can be incomplete,³ potentially instigating a treatment delay while further clinical information is gathered and an NMDT decision awaited.

The initial design and setup of the NMDT was aimed at patients requiring specialist intervention, and therefore commonly limited to a small group of patients presenting with a single metastasis and good prognosis from their systemic cancer.² In recent years there has been a rise in the incidence of CMs encountered in clinical practice because of improved diagnostic imaging techniques, a global increase in the incidence of primary cancer, and improved systemic treatments and overall survival.^4–6 As a result, there are increasing numbers of patients being referred to the NMDT with CM, some of whom may be suitable for treatment and others who will not benefit and thus are not appropriate for any intervention because of advanced disseminated disease.

The rationale for active intervention in CM was based on studies from the late 1990s showing a survival advantage and/or decrease from neurologic death conferred by a combined approach of neurosurgery or stereotactic radiosurgery (SRS) with adjuvant whole-brain radiotherapy (WBRT) in patients with oligometastatic disease.^7–10 A widely adopted prognostic scoring system used age, performance status, systemic disease burden, and presence of extracranial metastases (ECM) to stratify patients into 3 recursive partitioning analysis (RPA) classes with significantly different survival that were subsequently validated in various populations.⁷ More recent prognostic scoring systems have included the type of primary cancer and identified that the survival of patients with CMs varies significantly by diagnosis.¹¹ For each type of primary tumor, a disease-specific graded prognostic assessment (ds-GPA) score was derived to estimate survival.^11–14

However, there have been several recent changes in practice among specialists entailing a much more individualized approach in treatment decisions: First, there has been a move away from using WBRT, and SRS is now favored for multiple metastases as well as being used as treatment to the surgical cavity after resection.^15,16 Second, immunotherapy and targeted chemotherapy, such as checkpoint inhibitors, proto-oncogene BRAF V600E antibodies, or anaplastic lymphoma kinase inhibitors, have revolutionized the management of CMs from certain cancers such as melanoma and lung cancer.^17,18

Whereas NICE guidelines in 2006 recommended referral to the NMDT only for cases in which patients presented with solitary metastasis in good performance status with a prognosis warranting neurosurgical intervention or in cases where a referral was mandated to establish a diagnosis,² the newly published NICE guidelines from 2018 recommend referral for all CMs.¹⁹ Equally, treatment recommendations have been updated: Whereas formerly complete surgical removal of the solitary metastasis followed by postoperative WBRT was considered the mainstay of treatment, the new guidelines suggest a more complex approach, recommending: 1) surgery or SRS for solitary metastases with adjuvant SRS to surgical cavity in patients with 1 to 3 metastases, without adjuvant WBRT; 2) SRS/radiotherapy for patients with multiple metastases; and 3) WBRT only for patients who have not received surgery or SRS and who do not have non–small cell lung cancer.¹⁹

The aim of this study was to draw up a national picture of CM referrals and to assess whether decision making matches the changing landscape of metastasis management worldwide, and in light of the newly reformed NICE guidelines.²⁰

Furthermore, observational studies of CMs have been primarily of a retrospective nature and prospective studies have been restricted to a single center.^3,5,7,11 These limitations lead to inherent biases in practice and patient selection and may not reflect the current national practice to generate health economic models and allow future resource planning.²¹ Using prospectively collected data from multiple neurosurgical units, we aimed to assess the volume of CM referrals to the NMDT, the quality of referral information provided, and its impact on NMDT decision making. Thereby, the data presented in this study can be used as a baseline against which any future multicenter, randomized, controlled trials (RCTs) can be designed and adequately powered.

Materials and Methods

Study Design

A prospective, multicenter, observational study of CM management was conducted across 24 neurosurgical units in the United Kingdom and Ireland. Primary data collection took place over 4 months between November 2017 and February 2018 after an initial trial period at 1 center from September 2017 to October 2017 (see Supplementary Figures 1-3 for information on monthly recruitment and center participation, respectively). All adult patients (age ≥ 18 years) referred to the NMDT with CM were included in the study. The NMDT was composed of a variety of team members, including but not limited to the consultant neurosurgeon, neurologist, neuroradiologist, neuro-oncologist, neuropathologist, neuro-oncology clinical nurse specialists, occupational and speech and language therapists, physiotherapists, coordinators, and a neuropsychologist, where available. The study protocol was designed by the British Neurosurgical Trainee Research Collaborative (BNTRC)²² and approved by the Society of British Neurological Surgeons Academic Committee. The manuscript was written following the Strengthening the Reporting of Observational Studies in Epidemiology checklist.²³

Data Collection and Outcome Measures

Anonymized data were entered into Castor Electronic Data Capture, which is a secure online database complying with the Department of Health Information Governance policy and meeting the data security standards of the Information Governance Toolkit of the Health and Social Care Information Centre. The audit and clinical governance committee of each participating hospital approved the study protocol.

The following demographic and operative parameters were captured in the electronic case report form: age, sex, date of NMDT, presenting symptoms, KPS and Eastern Cooperative Oncology Group²⁴ performance status, status/location/diagnosis of primary disease, treatment of primary disease, presence of extracranial metastasis, positive/negative molecular markers of primary tumor, status of extracranial disease (local vs metastatic, controlled vs uncontrolled), cranial imaging undertaken, number/size/location of cranial metastases, delay of NMDT decision, treatment recommendation (“specialist” interventions as recommended by a dedicated neuro-oncology center (neuro-oncologist, neurosurgeon) located in a large tertiary referral unit: surgical resection, cerebrospinal fluid (CSF) diversion, SRS, cavity SRS; “nonspecialist” treatment as provided by a general oncologist: chemotherapy, immunotherapy, WBRT, local fractionated radiotherapy, best supportive care, other) and previous treatment of CM. RPA⁷ and ds-GPA¹¹ were calculated for all referred cases, providing the required information was completed.

Statistical Analysis

Descriptive statistics were used to characterize the patient population. Statistical analysis was performed using GraphPad Prism V7 and the Stata/IC v.15.1 statistical package. Chi-square test was used to assess the statistical significance of observed differences between cohorts undergoing specialist or nonspecialist treatment. Univariate logistic regression was used to explore the relationship between primary outcome (specialist vs nonspecialist treatment) and a set of predictors. Differences in the primary outcome (specialist vs nonspecialist treatment) between RPA classes I to III were represented with bar plots and analyzed with a chi-square test for trend.

Results

Patient Demographics, Performance Status, and Presenting Symptoms

In total 1048 patients were analyzed (Table 1) and 55.5% (n = 582) were female. Median age at referral was 65 years (range, 21-93 years) and the median number of referrals per weekly NMDT was 3 (range, 1-17). The most common presenting symptoms were motor deficit (30.1%, n = 315), headache (24.1%, n = 253), and confusion (17.9%, n = 188). A total of 6.8% of patients (n = 71) in our cohort presented with symptoms of increased intracranial pressure (ICP), and in 3.0% of cases (n = 31) CMs were found incidentally. KPS was 70 or greater in 54.8% (n = 564), less than 70 in 18.3% (n = 193), and not provided in 24.3% (n = 255).

Table 1.

Patient Demographics, Performance Status, and Presenting Symptoms

Variable	No. of Patients	%
Total	1048	100.0
Sex
• Female	582	55.5
• Male	466	44.5
Age, y
• <40	43	4.1
• 40-44	38	3.6
• 45-49	57	5.4
• 50-54	84	8.0
• 55-59	120	11.5
• 60-64	143	13.6
• 65-69	176	16.8
• ≥ 70	379	36.2
• NA	8	0.8
KPS
• 90-100	336	32.1
• 70-80	238	22.7
• 50-60	145	13.8
• 30-40	35	3.3
• 10-20	13	1.2
• NA	255	24.3
WHO performance status
• 0	187	17.8
• 1	369	35.2
• 2	184	17.6
• 3	81	7.7
• 4	22	2.1
• NA	205	19.6
Presenting symptoms
• Headache	253	24.1
• Motor deficit	315	30.1
• Speech deficit	128	12.2
• Visual deficit	67	6.4
• Seizure	115	11.0
• Confusion	188	17.9
• Screening	141	13.5
• Ataxia/LOC/falls	133	12.7
• Nausea/vomiting/increased ICP	71	6.8
• Weight loss/fatigue/lethargy	26	2.5
• Incidental finding	31	3.0
• Other/unknown	61	5.8

Abbreviations: NA, not available (unknown or not recorded); ICP, intracranial pressure; LOC, loss of consciousness; WHO, World Health Organization.

Pretreatment Characteristics: Primary Cancer

A total of 681 patients (65.0%) had a known primary diagnosis of cancer. The most common primary tumor locations were lung (36.5%, n = 383), breast (18.4%, n = 193), and melanoma (12.0%, n = 126) (Table 2). In 5.2% (n = 54) there was no extracranial disease. The primary tumor was controlled in 33.5% (n = 351), not controlled in 22.0% (n = 231), and this information was not provided in 39.3% (n = 412). A total of 44.6% (n = 467) of patients had extracranial metastases. The time interval between diagnosis of primary tumor and CM was 2 years or less in 33.7% (n = 353) and unknown or not recorded in 43.5% (n = 456). The status of markers of sensitivity to targeted chemotherapy in the primary cancer was unknown or not recorded in 71.3% of patients (n = 747).

Table 2.

Pretreatment Characteristics: Primary Cancer

Variable		No. of Patients	%
Total		1048	100.0
New diagnosis of cancer
• Yes		302	28.8
• No		681	65.0
• CUP		58	5.5
• NA		7	0.7
Location of primary
• Lung		383	36.5
• Breast		193	18.4
• Melanoma		126	12.0
• Upper GI tract		34	3.2
• Lower GI tract		58	5.5
• Kidney		49	4.7
• Prostate		13	1.2
• Genitourinary		46	4.4
• Multiple		23	2.2
• Other		43	4.1
• CUP/NA		80	7.6
Extracranial disease
• None		54	5.2
• Controlled	➢Primary site disease only	194	18.5
	➢Metastatic disease	157	15.0
• Uncontrolled	➢Primary site disease only	63	6.0
	➢Metastatic disease	168	16.0
• NA		412	39.3
Molecular markers
• Positive		216	20.6
• Negative		108	10.3
• NA		747	71.3
Time intervala, y
• ≤ 2		353	33.7
• > 2		239	22.8
• NA		456	43.5
Extracranial metastasis
• Yes		467	44.6
• No		536	51.1
• NA		45	4.3

Abbreviations: CUP, cancer of unknown primary; GI, gastrointestinal; NA, not available (unknown or not recorded).

^aTime between diagnosis of primary tumor and cerebral metastasis, where applicable.

Pretreatment Characteristics: Cerebral Metastasis

A total of 51.6% (n = 541) of patients were referred with a solitary CM. There were 31.0% (n = 325) who had 2 to 4 metastases (2 metastases: 18.2% [n = 191]; 3 metastases: 8.9% [n = 93]; 4 metastases: 3.9% [n = 41]), and 15.4% (n = 162) had 5 or more metastases (Table 3). Out of all patients referred, 14.7% (n = 154) had undergone previous surgery for removal of CM and were referred back to the NMDT for discussion of recurrent disease.

Table 3.

Pretreatment Characteristics: Cerebral Metastasis

Variable	No. of Patients	%
Total	1048	100.0
Number of brain metastases
• 1	541	51.6
• 2	191	18.2
• 3	93	8.9
• 4	41	3.9
• ≥ 5	162	15.5
• LMD	3	< 0.3
• NA	17	1.6
Sentinel location of lesions
• Frontal lobe	406	38.7
• Temporal lobe	79	7.5
• Parietal lobe	153	14.6
• Occipital lobe	96	9.2
• Cerebellum	203	19.4
• Brainstem	22	2.1
• Durally based	15	1.4
• Other	49	4.7
Size of sentinel metastasis, mm
• ≤ 30	637	60.7
• > 30	292	27.9
• NA	119	11.4
Cranial imaging
• CT head	635	60.6
• MRI head	873	83.3
• NA	13	1.2
Reason MRI not undertaken
• Contraindicated	17	9.7
• Patient unwilling	3	1.7
• Indicated but not performed before MDT	91	52.0
• No clinical indication	48	27.4
• NA	16	9.1
MRI sequences
• Gadolinium contrast	836	95.8
• Navigation sequence	378	43.3
• DTI	668	76.5
• DWI	66	7.6
• Spectroscopy	3	0.3
Prior brain surgery
• Yes	154	14.7
• No	891	85.0
• NA	3	< 0.3

Abbreviations: DTI, diffusion tensor imaging; DWI, diffusion-weighted imaging; LDM, leptomeningeal disease MDT, multidisciplinary team meeting; NA, not available (unknown or not recorded).

The most common sentinel locations of CM were the frontal lobe (38.7%, n = 406), the cerebellum (19.4%, n = 203), and the parietal lobe (14.6%, n = 153). A total of 83.3% (n = 873) of patients underwent MRI, and 60.6% (n = 635) of patients had a CT scan of the head prior to NMDT referral. Gadolinium contrast was administered in n = 836 (95.8% of MRI scans). In cases where MRI was not undertaken, the most common reason given was that the scan was indicated but not performed before the NMDT (52.0%, n = 91), followed by the second most common reason being that the referring team did not have a clinical indication to perform an MRI scan (27.4%, n = 48).

Treatment Recommendation

Specialist intervention (either SRS or surgical resection) was recommended for 52.6% (n = 551) of patients (Table 4). Specialist intervention was recommended for 67.5% (n = 365) of patients with a solitary metastasis, and for 38.2% (n = 186) of patients with multiple CMs. In particular, 48.6% (n = 158) of patients with 2 to 4 metastases and 17.3% (n = 28) of patients with 5 or more metastases were offered specialist intervention. The most commonly offered intervention was SRS alone (20.8%, n = 218), followed by surgical resection alone (18.7%, n = 196). A combination of (cavity) SRS and surgical resection was offered to 5.7% (n = 60). A combination of surgery or SRS with radiotherapy (WBRT or local fractionated radiotherapy) was offered to 1.7% (n = 18) and 0.5% (n = 5), respectively. Other surgical treatments offered to patients included a biopsy in 1.0% (n = 11), out of which 2 were for cancer of unknown primary and 5 for newly diagnosed patients, and a form of CSF diversion in 0.9% (n = 9).

Table 4.

Treatment Recommendation

Variable	1 CM	2-4 CMs	≥ 5 CMsa	NA	No. of Patients	%
Total	541	325	165	17	1048	100.0
Specialist intervention	365	158	28	0	551	52.6
• Surgical resection alone	163	31	2		196	18.7
• Surgical resection + SRS	8	27	2		37	3.5
• Surgical resection + SRS + cavity SRS	0	2	0		2	< 0.2
• Surgical resection + cavity SRS	21	0	0		21	2.0
• Surgical resection + chemotherapy/immunotherapy	4	2	0		6	0.6
• Surgical resection + WBRT/local fx Rx	12	5	1		18	1.7
• Surgical resection + CSF diversion	1	2	1		4	0.4
• SRS alone	126	74	18		218	20.8
• SRS + WBRT/local fx Rx	3	1	1		5	0.5
• SRS + chemotherapy/immunotherapy	14	5	1		20	1.9
• Biopsy alone	8	3	0		11	1.0
• Biopsy + SRS	0	1	0		1	< 0.1
• CSF diversion	5	2	2		9	0.9
• Clinic assessment to discuss surgery/SRS	0	3	0		3	< 0.3
Nonspecialist treatment only	165	147	133	1	447	42.7
• Chemotherapy	8	4	6		18	1.7
• Immunotherapy	3	2	3		8	0.8
• WBRT	20	40	54b	1	115	11.0
• Local fx Rx	11	4	1		16	1.5
• Oncology treatment NOS	13	22	14		49	4.7
• Best supportive care	68	62	50c		180	17.2
• Reimaging/surveillance	29	7	3		39	3.7
• Referral to other specialty	13	6	2		22	2.1
No MDT decision	11	20	4	16	51	4.9
• NA					45	4.3
• Indeterminate					6	0.6
Delay in MDT decision
• Yes					195	18.6
• No					767	73.2
• NA					86	8.2
Reason for delay (multiple)
• Imaging not available					102	52.3
• Insufficient information					53	27.2
• Awaiting further investigations/results					27	13.8
• Cancellation					2	< 0.2
• Wrong MDT					15	7.7
• Intentional delay					9	4.6
• Assessment					8	4.1
• Other					7	3.6

Abbreviations: CM, cerebral metastasis; CSF, cerebrospinal fluid; local fx Rx, local fractionated radiotherapy; MDT, multidisciplinary team meeting; NA, not available (unknown or not recorded); NOS, not otherwise specified (either WBRT or chemotherapy/immunotherapy or best supportive care); SRS, stereotactic radiosurgery; WBRT, whole-brain radiation therapy.

^aIncludes patients with leptomeningeal disease (LMD), n = 3.

^bIncludes n = 1 with LMD.

^cIncludes n = 2 with LMD.

In 42.7% (n = 447) of patients, the NMDT decision was to recommend nonspecialist treatment either in the form of active oncology treatment (chemotherapy 1.7% [n = 18], immunotherapy 0.8% [n = 8], or local fractionated radiotherapy 1.5% [n = 16]), or palliative treatment (WBRT 11.0% [n = 115] or best supportive care 17.2% [n = 180]).

In 18.6% (n = 195) of patients, there was a delay in the NMDT treatment recommendation given (median time to decision making after initial discussion in MDT was 11 ± 112 days) because of a lack of imaging (52.3%, n = 102), missing referral information (27.2%, n = 53), or waiting for further investigations or results (13.8%, n = 27).

Factors Influencing Neuro-Oncology Multidisciplinary Team Decision Making

Using univariate logistic regression, we explored the relationship between the primary outcome (specialist vs nonspecialist treatment recommendation) and independent predictors. We identified number of CMs, age, KPS, primary disease status, and extracranial disease as factors associated with NMDT decision making (Table 5, P < .001). Location of sentinel metastasis and histology of the primary tumor also showed a statistically significant association with NMDT decision making (P = .05 and P = .01, respectively). Factors that were not found to be associated with decision making were time interval to diagnosis, size of sentinel metastasis, prior brain surgery, preoperative neurological deficit, headache, and delay in NMDT decision (P > .05).

Table 5.

Factors Associated With MDT Decision Making Using Univariate Logistic Regression

Variable	Comparison	P
Number of cerebral metastases	Single vs multiple	< .001
Age, y	< 65 vs ≥65	< .001
KPS	< 70 vs ≥70	< .001
Primary disease status	Controlled vs uncontrolled	< .001
Extracranial disease	Brain metastasis only vs brain and other metastases	< .001
Sentinel location	Lobes/cerebellum vs brainstem/basal ganglia/other	.05
Sentinel size, cm	≤ 3 vs > 3	.11
Time interval, y	< 2 vs > 2	.93
Prior brain surgery	Yes vs no	.72
Histology of primary	SCLC vs TNBC	< .01
Preoperative neurological deficit	Yes (motor/speech/visual) vs no/missing	.09
Headache	Yes vs no	.10
Delay in MDT decision	Yes vs no	.28
MRI available	Yes vs no	> .001

Abbreviations: MDT, multidisciplinary team meeting; SCLC, small cell lung cancer; TNBC, triple-negative breast cancer.

Recursive Tree

With regards to RPA classes,⁷ only a small proportion of patients within our cohort were allocated to class I (n = 84, Fig. 1A). The majority of patients were either class II (n = 281) or class III (n = 190). RPA class I patients were managed surgically in the majority of cases (80.0%, n = 68), class II was managed either surgically (63.7%, n = 179) or nonsurgically (36.3%, n = 102, out of which WBRT was recommended in n = 43 and best supportive care in n = 30), and class III was managed nonsurgically in the majority of cases (66.8%, n = 127, out of which WBRT was recommended in n = 25 and best supportive care in n = 83).There was a statistically significant difference in surgical vs nonsurgical treatment between those 3 classes (chi-square_trendP < .0001; Fig. 1A and Supplementary Fig. 4).

Fig. 1.

Recursive Partitioning Analysis (RPA) of Study Patients and Treatment Recommendation per Disease-Specific Graded Prognostic Assessment (ds-GPA)

A, The recursive tree (adapted from Gaspar et al⁷) is a tool to classify patients into classes I to III. Patients with a KPS less than 70 are categorized as class III. Patients with a KPS of 70 or greater, controlled primary disease, age younger than 65 years, and no extracranial metastases (ECM) are classified as class I. All other patients are classified as class II. In our cohort the KPS was not available (NA) in approximately 25% of patients, chi-square_trend showed P less than .001. B, Patients were grouped into ds-GPA as previously described by Sperduto et al.45 The bar plots demonstrate the treatment recommendation (specialist/surgery vs nonspecialist/no surgery) per ds-GPA. There tended to be a higher proportion of recommended specialist treatment in patients with a higher ds-GPA score; however, these differences were not statistically significant with our data.

Validation of Disease-Specific Graded Prognostic Assessment

We applied ds-GPA classification for lung, melanoma, breast, renal, and gastrointestinal tract cancers (Fig. 1B). Overall, the proportion of recommendation for specialist treatment tended to be higher in patients with a high ds-GPA score and therefore longer expected median survival as compared to patients with a low ds-GPA score, but these differences were not statistically significant with our data. It is noteworthy that because of incomplete referrals, lack of KPS, molecular profile, and patient age there was a loss in numbers of patients, which was particularly evident in the breast and melanoma cancer group but also in gastrointestinal cancers, for which KPS was the only prognostic factor for median survival within this particular classification.

Discussion

Pattern of Cerebral Metastasis Referrals

There have been 3 large RCTs investigating the role of surgical resection in the treatment of solitary CM,^9,10,25,26 comparing surgical resection followed by radiotherapy vs radiotherapy alone. Two out of 3 RCTs found a statistically significant longer median survival and better quality of life in the surgical resection group. Two other large RCTs looked at the effect of SRS in combination with WBRT^15,27 in the management of single or multiple CMs and found that a combination of the 2 treatment modalities may show improved neurological function and intracranial tumor control; however, the combination does not show improved median survival. These findings were confirmed by a meta-analysis of 27 RCTs.²⁸

Current NMDT management is based on a combination of these studies with the evolving literature. Though WBRT has been the mainstay of treatment for decades, it has recently fallen out of favor because of its association with neurocognitive decline.¹⁶ Newer studies propose the use of SRS for multiple metastases and cavity SRS after surgical metastasis removal.^15,16

Additionally, advances in immunotherapy and targeted chemotherapy treatments offer alternatives to patients with a favorable mutation profile in melanoma and lung cancer.^17,18

In our cohort, 51.6% of patients were referred for treatment of a solitary metastasis. Within the subgroup of patients with multiple metastases, patients with 2 metastases were most commonly referred (18.2% of total) followed by patients with 5 or more CMs (15.5% of total). The change in practice reflects the fact that 38.2% (n = 186) of the patients referred with multiple metastases were recommended for specialist intervention, as compared to approximately 10% of patients in a single-center series of 1640 patients from 2013 to 2015.²⁹

Though treatment recommendation was limited to single CM in the former NICE guidelines of 2006, the newer NICE guidelines of 2018 give some recommendations regarding multiple metastases management; however, they lack any recommendation about surgical resection. Therefore, offering an intervention (surgery or SRS) in patients with multiple metastases remains entirely at the discretion of the NMDT and the treating surgeon or oncologist. In our cohort specialist treatment was recommended for 38.2% of patients with multiple metastases, suggesting evolving management strategies²⁷even before the publication of the 2018 NICE guidelines.

There have been some recent studies confirming an increase in the use of SRS alone for many patients with multiple CMs as a strategy to gain local control while minimizing cognitive effects associated with WBRT.³⁰ Though the benefit of surgical management of multiple CMs is currently lacking class I evidence, there are indications that surgery in these patients may be safe and beneficial to achieve intracranial tumor control, particularly to address large metastases, causing mass effect.³¹ Furthermore, a recent study suggests that redo surgery may also be a viable option in patients with recurrent CMs.³²

Referrals Requiring Specialist Intervention

In our cohort, 52.6% of patients required specialist intervention in the form of SRS or surgery. It is clear that the proportion of patients undergoing specialist treatment is negatively correlated with the number of metastases present at the time of referral.

Sills ³³ commented in 2005 on the evolution of treatment modalities in patients with CMs, due to improvements in surgical technique, using neuronavigation, presurgical mapping,³⁴ and intraoperative monitoring techniques, alongside diagnostic/therapeutic advances in the management of systemic cancers.^31,35 This may lead to a change in the role and timing of surgical resection as more and more (neo)adjuvant systemic therapies become available, making more patients eligible candidates for surgical resection. However, our cohort study confirmed that previously established factors^7,11 (such as age, KPS, number of CMs, presence of extracranial disease, and systemic disease status) still play a key role in specialist treatment recommendation in the form of either surgery or SRS, while stressing the importance of accurate disease staging at referral.^33,36–41 One factor that could not be analyzed because of lack of data is the influence of molecular marker status on NMDT decision making, which may be crucial in some cancer subtypes to make the best decisions.

In fact, after categorizing our cohort into groups based on the recursive tree, 2 main things can be observed: First, a significant proportion of patients (18.3%) are referred with a KPS less than 70 and therefore per se fall into the category of patients with poor median survival⁷ and are therefore poor surgical candidates (albeit ~30% of those had specialist treatment recommended, suggesting that there is a necessity to discuss these patients in the NMDT). Second, there was a large proportion of patients (24.3%) for whom the KPS was not provided by the referring team. Increasing compliance with KPS reporting at referral would therefore help streamline decision making at the NMDT.

We found no evidence of an association between the following prognostic factors⁷ and NMDT decision making in our cohort: prior brain surgery, time interval between primary and secondary tumor diagnosis (before or after 2 years), neurological dysfunction, and/or headache at presentation. The fact that having undergone prior brain surgery for removal of metastasis excluding further specialist intervention within our data supports the idea of redo surgery as an option that can have good outcomes in selected patients.³⁴

Delay in Multidisciplinary Team Decision Making

In approximately one-fifth of patients referred (18.6%), there was a delay in NMDT decision making. The most common reasons given were incomplete referral information provided, lack of imaging availability for review, and/or awaiting further investigations or results from the referring team. This may lead to an increase in NMDT workload because those factors are considered essential for the decision-making process. Nonetheless, the fact that the NMDT decision was delayed did not influence the outcome of the treatment recommendation given (Table 5, P = .278). Whether the delay in offered treatment has a negative impact on patient survival will have to be assessed in future studies.

Potential solutions would include reiterating to referring teams the importance of all the information required, and identifying and supporting those teams that repeatedly send incomplete referrals. New streamlining pathways could also be established including an emphasis on a uniform national proforma in which data (including molecular profiles) is collected continuously, perhaps even capturing national outcome data. A further advantage of this would be that all required data would be readily available and could be shared between all specialties (general practitioners, emergency departments, oncologists, neurosurgeons, etc).

Validation of Recursive Partitioning Analysis and Disease-Specific Graded Prognostic Assessment

The use of RPA and ds-GPA has been previously validated.⁴² More recently, molecular subtypes of tumors have also been taken into account, first in breast⁴³ and then in lung cancer.⁴⁴ Overall, our data showed that the better the RPA class⁷ (ie, RPA class I) the more likely the patient was to have specialist treatment recommended. Though there tended to be a greater chance of specialist treatment with a higher ds-GPA score,^11,45 we did not find a statistically significant association with our data.

One of the reasons for the compliance rate falling short of 100% could be the recent developments in surgical techniques leading to a wider variety of patients being considered for such treatments. A recent study of 71 patients at a single institution showed that the actual survival outcome exceeded expected outcome significantly in a well-selected cohort of patients.⁵ This remains to be confirmed in a larger patient population. Another reason could be that more surgery is offered to the elderly, as an increasing number of otherwise fit patients are referred in an aging population.²⁹

There have been efforts to develop new stratification tools such as the Barnholtz-Sloan index,⁴⁶ Score Index for Radiosurgery, and Basic Score for Brain Metastases, among others^6,47,48 to guide NMDT decision making for this heterogeneous cohort of patients. These have not been widely adopted into clinical practice for a number of reasons, presumably because most of these scores are based on survival data alone without considering other important factors such as quality of life and tumor recurrence. Other reasons may be related to the constant evolution of molecular profiling and new therapeutic targets.^18,49 Overall, population-based studies are not always as good at predicting individual outcome, and it is evident that CM management has become very complex and a much more individualized approach is being applied. In the near future, one of these may be complemented by the use of imaging as a potential biomarker.⁵⁰

Data Generalizability and Limitations of This Study

The primary advantage of this study is its multicenter nature, allowing for a large sample size. Three-quarters of neurosurgical centers in the United Kingdom and Ireland participated in this cohort study, which gives a reflection of national management of CM referrals. Regional homogeneity of the referred patient population and NMDT treatment recommendation provided is of vital importance to plan future RCTs, inform health policy makers (including NICE), generate health economic models, and assist in national resource allocation. In the future, we would welcome a prospective national database for CM referrals that captures national outcome data.

One of the limitations of this study has been that some of the referral information has been largely incomplete or missing as a whole. This limitation lies within the nature of this study and can be largely attributed to lack of information at the time of referral and does not reflect on the quality of data entry.

Furthermore, although SRS to the resection cavity is supported by NICE if there is residual disease documented by postoperative MRI, this may not be recommended at the initial NMDT. Therefore, a proportion of patients will have had cavity SRS without this being captured in this study.

Conclusions

The development of new NICE guidelines will lead to an increase in NMDT workload. Our prospective study identified a delay in NMDT decision making for approximately 1 in 5 patients. Specialist intervention was offered to 67.5% of patients with a single CM and 38.2% of patients with multiple CMs, hence confirming a national change in culture of referral and treatment patterns, including a general trend away from adjuvant WBRT and specialist treatment being more frequently offered for patients with multiple CMs.

Funding

The post of CB is partly funded by the National Institute for Health Research (NIHR) Biomedical Centre based at Guy’s and St. Thomas’ NHS Foundation Trust and King’s College London. The views expressed are those of the authors and not necessarily those of the National Health Service, the NIHR, or the Department of Health.

Conflict of interest statement

None declared.