Background

Brain metastases occur in 20-40% of cancer patients during the course of their disease, and are thus a frequent cause of morbidity and mortality. Based on computed tomography (CT) scanning, about half of these patients have solitary lesions, although this falls to a quarter to a third with magnetic resonance imaging (MRI) brain due to its greater sensitivity, and the true rate may be as low as 10-20%.1,2,3 Although the median survival with conventional treatment viz. corticosteroids and whole brain radiotherapy (WBRT) is 3-6 months, subsets of patients do realise significantly longer survival. Retrospective analyses have identified several strongly predictive prognostic factors. For example, Gaspar et al. in 1997 described three recursive partitioning analysis (RPA) classes: class 1 patients had Karnofsky Performance Status (KPS) >= 70, age < 65, controlled primary and no extracranial metastases, class 3 patients had KPS < 70 (unable to care for themselves) and all others were in class 2. The median survivals for these patients who were treated (essentially) with WBRT were 7.1, 4.2 and 2.3 months for RPA classes 1, 2 and 3, respectively.4 Subsequently, several other prognostic indices were developed. Most recently, the Diagnosis-Specific Graded Prognostic Assessment of Sperduto et al. added primary site and number of brain metastases to Gaspar's original prognostic factors, with median survivals for the various groups ranging between approximately 3 and 19 months.5 Variations of this magnitude due to patient and tumour characteristics mean that any interventions need to be compared in a randomised fashion in order that imbalanced selection factors do not overwhelm any potential differences in outcome due to treatment. For this reason, the following literature review will focus on randomised data.

Since the landmark study of Patchell et al. was published in 1990,1 numerous other randomised trials have investigated strategies for patients most likely to benefit from aggressive local treatment viz. those of good performance status with single or few (oligo-) brain metastases. These trials have assessed the relative roles of surgery (S) and stereotactic radiosurgery (SRS) (highly focused single fraction radiotherapy utilising stereotactic technology), with or without WBRT. The median survivals reported for the relatively favourable prognosis patients enrolled in these trials are typically 6-11 months.1,2,3,6,7,8,9,10,11,12,13 The purpose of this Commentary is to summarise the findings of these randomised trials as an appropriate evidence base for the management of patients with solitary brain metastases. As some of the trials included patients with up to three or four metastases, conclusions are drawn where possible for this scenario also. None of the trials included protocol systemic therapy, nor have they so far reported relative costs, issues therefore beyond the scope of the present work.

Methods

Published randomised trials on S or SRS with or without WBRT for solitary or oligo-brain metastases were identified via a search of MEDLINE, EMBASE and Cochrane databases using various combinations of the search terms solitary/single/oligo-brain/cerebral metastasis/metastases, surgery/neurosurgery/microsurgery, radiosurgery/stereotactic radiosurgery, whole brain radiotherapy/radiation therapy/irradiation and randomised/controlled trial.

Results

The search identified 14 trials (3 in Abstract form only) which have addressed four main questions and two other related issues as follows:

Is there an added benefit of surgery to WBRT?

There have been three published randomised trials comparing WBRT + S versus WBRT alone. Patchell et al. found increased median survival (40 vs. 15 weeks, P < 0.01), decreased local recurrence (20% vs. 52%, P < 0.02) and increased time maintaining KPS >= 70% (38 vs. 8 weeks, P < 0.005) with the addition of S to WBRT.1

Noordijk et al. (previously reported in Vecht et al.) also found a survival benefit with S (median 10 vs. 6 months, P = 0.04), but this benefit was confined to patients with 'inactive extracranial disease' defined as stable, controlled or no evidence during the previous 3 months (12 vs. 7 months, P = 0.02, compared with 5 months in both arms for patients with active extracranial disease) or to patients aged 60 or less (19 vs. 9 months, P = 0.003, compared with 6 vs. 5 months for patients older than 60). There was also a borderline improvement in functionally independent survival in the S arm (median 7.5 vs. 3.5 months, P = 0.06). Local brain recurrence rates were not reported in this trial, although death due primarily to central nervous system (CNS) cause was similar in each arm (35% vs. 33% where this could be determined clinically).6,7

On the other hand, Mintz et al. found no improvement in survival from the addition of S to WBRT (5.6 vs. 6.3 months, P = 0.24), not even for the better prognosis patients free from extracranial metastases. Nor was there any difference between the arms in the time maintaining KPS >= 70. Again, local brain recurrence data were not reported, and death primarily due to CNS cause was similar in each arm (15% vs. 28%, P = 0.3).8

The somewhat inconsistent results from these three trials could be due to many reasons. The sample sizes were small (48, 63, 84 patients, respectively). The extent of baseline investigations differed significantly (all of Patchell et al.'s patients had MRI brain and all but 6 of 23 in the WBRT alone arm had histological confirmation; in the Noordijk et al. trial MRI was optional and biopsy was not performed; Mintz et al.'s patients had CT alone and biopsy only if 'the diagnosis was uncertain on CT'). Hence, different proportions of patients randomised to the WBRT alone arms could have had other than true solitary metastases. Notably, six of 54 patients (11%) initially registered in Patchell et al.'s trial were excluded after biopsy or excision confirmed alternative diagnoses (glioma, infective or inflammatory). Furthermore, the trials had different eligibility criteria (Mintz et al. allowed KPS >= 50%, Patchell et al. KPS >= 70%, Noordijk et al. WHO Performance Status <= 2) and different proportions of patients with systemic involvement (Patchell et al. 79%, Mintz et al. 79%, Noordijk et al. 32% with active extracranial disease).1,7,8 In short, the trials had different patient mix.

In summary, at least for the subset of good prognosis patients with resectable solitary brain metastases, S prolongs maintenance of performance status, local control and survival relative to WBRT alone. The evidence of benefit for poor prognosis patients appears weak. At the time of writing, there are no published randomised data evaluating the addition of S to WBRT for multiple brain metastases.

Is there an added benefit of radiosurgery to WBRT?

There have been two published randomised trials comparing WBRT + SRS versus WBRT alone, that is, assessing SRS 'boost'.14 The Radiation Therapy Oncology Group (RTOG 95-08) accrued 331 eligible patients with one to three brain lesions. Radiosurgery was reported to improve survival for patients with solitary metastases (median 6.5 vs. 4.9 months, P = 0.039, although it is noted that the survival advantage was borderline on multivariate analysis, P = 0.053), but it did not improve survival for multiple brain metastases (median 5.8 vs. 6.7 months, P = 0.98). Survival benefit from SRS was also explored for other subgroups (e.g. RPA class 1, largest tumour > 2 cm), but none of these factors were pre-stratified nor were they significant on adjusted P-value or on multivariate analysis. The WBRT + SRS arm overall had improved 1-year local control for centrally reviewed cases (82% vs. 71%, P = 0.01) with a higher rate of stable or improved KPS (43% vs. 27%, P = 0.03) and decreased steroid use at 6 months, but there were no differences between the arms in neurological (or other) causes of death for both the single and multiple metastases groups.9

Kondziolka et al. terminated their randomised trial of WBRT + SRS versus WBRT after slow accrual of only 27 patients when an interim analysis showed 1-year local brain failure rates of 8% versus 100%, respectively (P = 0.0016). There was no difference in survival with the radiosurgery boost (P = 0.22); maintenance of performance status was not reported. However, this trial was limited to patients with two to four brain metastases only.3

For completeness, mention is made of an Abstract from 2000 which reported 96 patients with one to three brain metastases randomised to three arms: WBRT versus SRS versus WBRT + SRS. However, interpretation was complicated by the fact that 51 patients had resection of 'large, symptomatic lesions' before randomisation. This study does not appear to have subsequently been published and will not be further considered here.15

In summary, for solitary brain metastases, the addition of SRS to WBRT modestly improves survival, local control and maintenance of KPS relative to WBRT alone. There is no survival benefit for multiple brain metastases although local control and maintenance of performance status are improved with SRS.

Is there an added benefit of WBRT to surgery or radiosurgery?

That aggressive local treatment (S or SRS) may confer some benefit over WBRT alone raises the question as to whether the adjuvant WBRT itself is necessary. Five trials examining this issue have so far been reported. Patchell et al. randomised 95 solitary brain metastases patients to S + WBRT (50.4 Gy in 28 fractions) versus S alone, finding that the addition of WBRT significantly decreased the rate of brain recurrence at any site (18% vs. 70%, P < 0.001), local recurrence (10% vs. 46%, P < 0.001) and neurological death (14% vs. 44%, P = 0.003), but did not affect survival (median 11 vs. 10 months, P = 0.39), or median time maintaining KPS >= 70 (8.5 vs. 8 months, P = 0.61). This was because patients treated with upfront WBRT tended to die instead from systemic disease (84% vs. 46%, P < 0.001).2

The Trans Tasman Radiation Oncology Group (TROG 98.05) planned to randomise patients with one or two brain metastases treated with either S or SRS to WBRT (initially 36 Gy in 18 fractions, later 30 Gy in 10 fractions) versus no WBRT, but the study was abandoned due to slow accrual after only 19 patients (17 S, 2 SRS), all with solitary lesions. Nonetheless, the conclusions were in line with Patchell et al.'s. There was no difference in overall survival (median 9.2 vs. 6.2 months, P = 0.99) or time to deterioration in WHO Performance Status to >=2 (P = 0.8), and a trend in time to CNS relapse favouring the WBRT arm (P = 0.12).10

In a similar study for patients with one to three brain metastases, the European Organisation for Research and Treatment of Cancer (EORTC) randomised 359 patients who were to be treated with either S (160) or SRS (199) to WBRT (30 Gy in 10 fractions) versus no WBRT. At the time of writing, this has been reported in Abstract form only. Again, there was no difference in median survival (11 months each arm) and the intracranial progression rate was significantly decreased by the addition of WBRT (31% vs. 54% at 24 months, P < 0.0001). As a result, consistent with Patchell et al.'s trial, there were fewer neurological deaths in the WBRT arm (25% vs. 43%, P-value not given) but no difference in duration of functional independence (median 9.8 months in each arm, P > 0.5).13

The outcomes were also very similar for Aoyama et al. who randomised 132 of a planned 178 patients with one to four brain metastases to SRS + WBRT (30 Gy in 10 fractions) versus SRS alone (SRS doses in the combined arm were reduced by 30% due to toxicity concerns). There was no difference in median survival (7.5 vs. 8.0 months, P = 0.42), but decreased 1-year intracranial relapse with WBRT at any site (47% vs. 76%, P < 0.001) and at the index site (11% vs. 27%, P = 0.002), although in this case, it did not translate to a difference in preservation of neurological function or rate of neurological death (23% vs. 19%, P = 0.64). As with the other trials, there was no significant difference in preservation of systemic function (KPS >= 70%).11

Most recently, Chang et al. at the MD Anderson Cancer Centre randomised 58 of a planned 90 patients with one to three metastases to SRS + WBRT (30 Gy in 12 fractions) versus SRS alone. This trial was closed early on the basis of neurocognitive stopping rules (see Neurocognitive function). Again, upfront WBRT decreased 1-year intracranial relapse (27% vs. 73%, P = 0.0003) and local relapse (0% vs. 33%, P = 0.012). Although there was similar preservation of systemic function (median KPS 70 at 4 months with WBRT vs. 80 without WBRT) and a similar rate of neurological deaths (25% vs. 27%, P = 0.15), WBRT was associated with more systemic deaths (57% vs. 33%, P = 0.013). Contrary to the other four trials, there was a significant survival difference between the arms (median 15.2 months without WBRT vs. 5.7 months with WBRT, P = 0.003). This was an unexpected finding, the reasons for which are unclear.16

In summary, when added to S or SRS, WBRT decreases local and distant brain relapse but does not improve survival (worse in the Chang et al. trial only) or maintenance of functional status. Reported causes of death (neurological vs. systemic) have been somewhat inconsistent in these trials.

Are surgery and radiosurgery equivalent?

The question of whether S and SRS produce similar outcomes has proven very difficult to address in the randomised setting. In fact, the comparison is directly relevant for only a small proportion of brain metastases patients, as few are truly suitable for both S and SRS. Many are precluded from S by site (e.g. brainstem, deep/eloquent supratentorial location) or medical comorbidities/poor performance status, while SRS is unsuitable for lesions larger than 3-4 cm in diameter, where tissue is necessary for diagnosis, that is, no systemic disease to biopsy (given that the trend is for excision biopsy in this setting) or in the presence of life-threatening raised intracranial pressure. In addition, when faced with the choice, some patients decline S because they wish to avoid an invasive procedure and/or general anaesthetic, and others decline SRS because they want the lesion excised, while some want neither. The first attempt to mount a randomised trial based at the Harvard Joint Centre for Radiation Therapy was abandoned in 1995 due to slow accrual after only seven patients.

A multicentre German solitary brain metastases study by Muacevic et al. was terminated early due to slow accrual after 64 of a planned 240 eligible patients. However, this trial was not a pure comparison of the treatments because adjuvant WBRT was included in the S but not the SRS arm. They found similar survival (median 9.5 vs. 10.3 months, P = 0.8), 1-year neurological death (29% vs. 11%, P = 0.3) and 1-year local control (82% vs. 97%, P = 0.06) with S + WBRT and SRS alone, respectively. There was a significantly increased 1-year distant brain relapse rate in the SRS alone arm (3% vs. 26%, P < 0.05), consistent with the absence of upfront WBRT.12

In another prematurely closed trial, Roos et al. compared S with SRS for 21 solitary brain metastases patients at a single centre (originally planned to be a multicentre study). However, unlike the Muacevic et al. trial, adjuvant WBRT was included in each arm, hence a valid comparison of the two local interventions. Of note, 18 patients deemed eligible declined randomisation for the reasons given above. Acknowledging the low statistical power, there were no significant differences in overall survival, failure-free survival or maintenance of systemic or neurological functional status. Both strategies were very effective in preventing local relapse, only one such event being recorded (Assoc Prof Daniel Roos, Dr Jennifer Smith, Ms Sonya Stevens, submitted, 2010).

Lang et al. also conducted a single centre direct comparison at the MD Anderson Cancer Centre, randomising 59 patients over 8 years to S versus SRS neither with WBRT, together with a parallel cohort of 155 patients who declined randomisation. Published in Abstract form only at the time of writing, the separate randomised results are awaited.17

Clearly the data on this question are limited at present, with three different approaches to adjuvant WBRT in the completed studies. Although it would appear that S and SRS may offer similar survival and local control, very few patients with solitary brain metastases are truly suitable for both options.

Quality of life

Quality of life (QOL) is notoriously difficult to study in the palliative setting. This is especially so for brain metastases patients who may have rapidly declining performance status and cognitive function rendering completion of QOL forms problematical, although there is some evidence that carers can validly serve as proxy raters.18

At the time of writing, QOL has only been reported in four of the above trials. Mintz et al. used the five-domain Spitzer QOL index finding no difference during the first and second 3-month periods after WBRT + S versus WBRT alone.8 The other three trials used the EORTC QLQ-C30 core questionnaire and QLQ-BN20 brain cancer module. The Trans Tasman Radiation Oncology Group found no difference in the two global QOL questions (overall health and overall QOL) at 2 or 5 months for the WBRT versus no WBRT arms. There were too few patients for further analysis to be appropriate.10 Muacevic et al. reported improvement in two (only) of seven QOL domains at 6 weeks for SRS relative to S + WBRT (P < 0.05), but the absence of WBRT in the former arm renders this finding difficult to interpret. In any event, the difference was lost by 6 months.12 Finally, Roos et al. found no differences in QOL at 2 months between S + WBRT versus SRS + WBRT, but again, further analysis was precluded by the small number of surviving patients (Assoc Prof Daniel Roos, Dr Jennifer Smith, Ms Sonya Stevens, submitted, 2010).

Clearly the currently available data are quite limited, but there appear to be no major differences in QOL between the various treatment options.

Neurocognitive function

The potential of WBRT to cause late neurotoxicity is usually referenced in the literature to the paper of De Angelis et al. in 1989. However, this was based on only 12 patients developing dementia (out of many hundreds of patients receiving WBRT at several centres), 10 of whom were treated with hypofractionated schedules (large doses per fraction) which are no longer in standard use.19 The above trials used a wide range of WBRT schedules, but most commonly 30 Gy in 10-12 fractions, regarded as conventional and 'safe' by most radiation oncologists. Nevertheless, the risk of at least subtle impairment of higher mental function from WBRT needs to be weighed against its consistently reported benefit in reducing local and distant brain recurrence adjuvant to S or SRS (see Is there an added benefit of WBRT to surgery or radiosurgery?).

Three of the above trials assessed changes in mini-mental state examinations (MMSE). No differences between the WBRT versus no WBRT arms were found by either Roos et al. or Aoyama et al., the latter concluding that 'control of brain tumours is the most important factor in stabilising neurocognitive function for most brain metastatic patients'.10,20 Andrews et al. also confirmed no difference in MMSE from the addition of SRS to WBRT versus WBRT alone.9

However, the limitations of the MMSE alone are well recognised. More thorough neurocognitive assessment is resource-intensive and difficult to conduct in the population under discussion. Not surprisingly, therefore, randomised data are scarce. Only one of the trials has so far reported using a battery of formal neurocognitive tests. Chang et al. found that patients treated with SRS + WBRT were 'significantly more likely to show a decline in learning and memory function (mean posterior probability of decline 52%) at 4 months than patients assigned to receive SRS alone (mean posterior probability of decline 24%)', as assessed by the Hopkins Verbal Learning Test - Revised. This was the reason why the trial was stopped early (see Is there an added benefit of WBRT to surgery or radiosurgery?), and the difference was reported to persist at 6 months. However, differences for the other four administered neurocognitive tests were much less striking and indeed, not all favoured the SRS alone arm (web appendix), but they may have been underpowered because of early closure. Contrary to Aoyama et al.,20 Chang noted that this differential neurocognitive decline occurred despite a significantly higher rate of brain relapse in the SRS alone arm. This was explained by close surveillance detecting relapse early, thereby allowing intervention whilst still asymptomatic, but it was at the expense of a much higher rate of salvage therapies in the no WBRT arm (26/30 = 87% of patients) than in the SRS + WBRT arm (2/28 = 7%).16 Again, Chang et al.'s findings conflict with the previous literature that suggests 'progression of brain metastases is a much greater cause of neurocognitive dysfunction than WBRT' (well discussed in Sperduto et al.).5 It is possible that the unexpected survival difference in favour of SRS alone reported in Chang et al.'s trial may have influenced the neurocognitive comparison between the arms, and there would appear to be a need for confirmatory randomised data before the conclusions should be accepted as definitive.

In the meantime, patients need to be made aware of the potential neurocognitive risks associated with both WBRT and with relapse without it, in order that they can make an informed decision on adjuvant WBRT at the time of S or SRS. Some patients, perhaps many, may prefer to minimise the need for subsequent intervention by having WBRT upfront.

Discussion

This review of randomised trials on aggressive treatment of solitary and oligo-brain metastases has revealed some significant limitations. Most have had small sample sizes (<100 patients),1,2,3,6,7,8,10,12,15,16,17* and many have been terminated early as a consequence of slow accrual, interim analysis, or both3,8,10,11,12,16* (note that Vecht/Noordijk et al. did not report the planned sample size6,7). Similarly, a well-publicised currently accruing American College of Surgeons Oncology Group/North Central Cancer Treatment Group SRS +/- WBRT trial has also amended its original sample size from 480 to 180 (ACOSOG Z0300/N 0574). Although the brain metastases population is enormous, few patients have proven amenable to randomisation for aggressive intervention strategies. Consequently, only large cooperative groups have been able to conduct well-powered studies.9,13 It is also important to acknowledge that the trial results are in any event only applicable to a relatively small proportion of brain metastases patients. All but two of the above trials excluded patients with poor performance status defined as KPS < 70 or ECOG/WHO > 2 (and only 21 such patients were registered on those two trials8*). Furthermore, many are unsuitable due to contraindications to S or SRS (see Are surgery and radiosurgery equivalent?), or they decline aggressive intervention.

The pitfalls of imaging diagnosis of solitary (presumed) brain metastases have been alluded to in the section Is there an added benefit of surgery to WBRT?. Thus a small proportion of trial patients will have the wrong diagnosis. This is more problematical with SRS where there is usually no pretreatment biopsy of the brain lesion(s). Imaging is also fallible in follow-up where postoperative and post-SRS changes (particularly radionecrosis) can be difficult to distinguish from relapse. Because autopsy is uncommonly performed in the setting of disseminated malignancy, pathological confirmation of recurrent brain metastases and/or radionecrosis is often not available, even on trial. Hence, quoted local control rates are subject to error. These issues need to be borne in mind when interpreting research on solitary or oligo-brain metastases.

While the side-effects of S and SRS are obviously different, and may well influence patient choice, they are both very safe procedures with reported acute and late grade 3/4 toxicities typically <5% in the above trials and no statistically significant differences between the arms. This was the case both in the two head-to-head comparisons which reported toxicity12* and in all the single modality trials.

Relative to the large survival differences attributable to patient and tumour factors alone4,5 (see Background), it can be argued that the contribution from aggressive local intervention for brain metastases (S or SRS) is modest indeed. The benefit from S is frequently overstated, or at least oversimplified, in the literature by preferential reference to the positive Patchell et al. trial (949 citations, ISI web of science, accessed 13 July 2010)1 at the expense of the restricted benefit reported by Vecht/Noordijk et al. (311 and 266 citations, respectively)6,7 and the negative result of Mintz et al. (212 citations).8 Of significance, the survival benefit in Patchell et al.'s trial for the S + WBRT arm (median 40 weeks = 9.2 months) was in the context of an atypically poor survival for the WBRT alone arm (15 weeks = 3.5 months) and a sample size of only 48 patients. Similarly, the improvement in median survival from SRS reported by Andrews et al. for patients with solitary metastases was only 1.6 months (6.5 vs. 4.9 months).9

Brain metastases trials registered on the ClinicalTrials.gov website include several ongoing and recently closed studies asking similar questions to those discussed above. There are also many other issues currently being evaluated in phase II or III trials. These include comparison of conventional conformal external beam radiotherapy boost versus SRS, WBRT with integrated boost to metastases (helical tomotherapy), the incorporation of chemo- or biological therapy with brain irradiation, SRS as a boost to the surgical excision site, the roles of brachytherapy (internal radio-isotope treatment) and (fractionated) stereotactic radiotherapy, adjuvant radiotherapy to less than whole brain (e.g. posterior fossa only) and strategies for specific primary sites (e.g. melanoma). Thus the management of solitary (or few) brain metastases remains very much an active area of research, recognition that many questions remain controversial or unanswered at this time.

Conclusions

This review has identified many limitations of the currently available randomised data on S or SRS for solitary and oligo-brain metastases. The results therefore need to be interpreted with caution. However, it is reasonable to conclude that for solitary brain metastases, both S and SRS improve local control, maintenance of performance status and survival relative to WBRT alone. The survival benefit from S may be limited to (or at least most pronounced for) good prognosis patients (especially those without active extracranial disease). Limited data suggest similar outcomes from S and SRS, but few patients are truly suitable for both options. The absolute survival benefit for most patients is modest.

For multiple (2-4) brain metastases, there is no survival benefit from SRS, although it does improve local control and functional outcome relative to WBRT alone; there is no randomised evidence for S.

When added to S or SRS, WBRT improves local and distant brain control, but not survival (one conflicting trial suggests WBRT worsens survival when added to SRS16). The neurocognitive risk : benefit ratio of WBRT remains controversial.

Randomised QOL data are sparse, but no major differences between the treatment options have so far been reported.

References