For many years, surgical resection of brain metastases was considered a form of palliative therapy only, having most benefit for patients with radioresistant tumors, such as colon carcinoma, melanoma, or renal cell carcinoma.  More recently, prospective studies have demonstrated that, in appropriately selected patients, surgical treatment can effectively prolong survival in patients with brain metastases. Complete surgical removal of a brain metastasis leads to immediate elimination of a mass compressing surrounding brain structures or causing blockage of the CSF flow, and removal of the source of perifocal edema. Surgery is particularly useful for patients with large (> 3 cm) lesions, especially those in the posterior fossa. Surgical patients may also benefit from a taper of steroids in the postoperative period, thereby reducing the risk of potential complications associated with their use.

Surgical resection of metastases is also valuable to confirm or define the diagnosis. In patients with no known primary cancer it may be difficult to differentiate metastatic disease from a primary brain tumor based on clinical and radiographic grounds alone. Indeed, even in patients with a known primary cancer, a newly diagnosed brain mass can turn out to be a primary brain tumor or other nonmetastatic disease in approximately 5% of patients.  Even in an era of widespread use of novel imaging techniques (eg, magnetic resonance spectroscopy, positron emission tomography, and so on), surgery remains the only treatment modality that provides actual tissue diagnosis.

The benefit of surgery in the treatment of a single brain metastasis has been demonstrated in two prospective phase III studies. Patchell  prospectively randomly assigned 48 patients with a single brain metastasis to surgical resection followed by whole-brain radiotherapy (WBRT; 36 Gy, 25 patients) or WBRT only (36 Gy, 23 patients). The median survival of patients in the surgical group was prolonged significantly compared with the radiotherapy-only group (40 v 15 weeks; P < .01), as was length of functional independence (38 v 8 weeks; P < .005). Similarly, Vecht randomly assigned 63 patients to surgery and WBRT (40 Gy, 32 patients) versus WBRT only (40 Gy, 31 patients) and observed a significant prolongation in survival (10 v 6 months; P = .04) and a trend toward improved maintenance of functional independence (7.5 v 3.5 months; P = .06). One negative study (Mintz) randomly assigned 84 patients to surgery and WBRT (30 Gy, 41 patients) versus WBRT (30 Gy, 43 patients) and found no significant difference in survival (5.6 v 6.3 months; P = .24) or maintenance of functional independence. It should be noted, however, that 73% of patients in this study had poorly controlled extracranial disease, unequal distribution of primary pathologies between the groups (ie, greater proportion of radioresistant colorectal cancer in the surgery group and radiosensitive breast cancer in WBRT group), and nonuniform calculation of survival times.

van der Ree  reported that surgical resection of brain metastases, particularly in the posterior fossa, may cause leptomeningeal dissemination of the tumor. In their series, 33% of patients developed leptomeningeal metastasis 2 to 13 months after surgery, which included six of the nine patients operated on for posterior fossa metastasis. Some authors, therefore, suggest an en bloc removal of metastatic lesions.

The justification for the use of surgical resection for the treatment of multiple brain metastases is less clear than it is for single or solitary brain metastasis. Uncontrolled, retrospective studies suggested that patients with multiple brain metastases did not benefit from surgery, compared with the historical results with WBRT alone.   These studies included patients with advanced systemic disease, or patients in whom only a fraction of the total number of metastases were treated with surgery. A significantly longer survival was found in patients who underwent removal of all metastases than in patients in whom at least one lesion was not resected (14 v 6 months). The conclusion of this study was that removal of all tumors in patients with multiple metastases may have survival rates comparable to that observed after resection of a single metastasis. However, it is also possible that patients who have tumors restricted to locations where they can be removed safely have a better prognosis than do patients who have tumors that are in the brainstem or that involve functional cortex, where the risks of surgery are unacceptably high. It is important to recognize that no randomized prospective studies have been conducted to address the question of whether surgery provides benefit for patients with multiple brain metastases.

The primary goals of surgical resection are to achieve a gross total resection of all tumor tissue and to minimize the risk of producing new neurologic deficits. Risk of neurologic injury related to surgery is determined mostly by tumor location; lesions at the surface of the brain can most often be removed completely with small risk of injury, whereas subcortical tumors, particularly those that are within or adjacent to anatomically defined functional areas, may or may not be safely removed, and decision making must be individualized to the patient.

Recent technologic advances in surgical adjuncts have provided neurosurgeons with new tools with which to expand the indications for which tumors can be considered resectable. These tools include computer-assisted image-guided surgery, intraoperative MRI, intraoperative ultrasound, functional MRI, and intraoperative electrocorticography. For image-guided surgery, preoperative images (including various sequences of regular MRI, functional MRI, or CT) are transferred to a computer workstation (navigation system) where they are coregistered with radiographic markers on the patient’s head and a reference device that is fixed to the head-holder and that can be tracked throughout the procedure by the navigation system. The surgeon then can easily identify the tumor and normal structures on the navigation system planning station. The high degree of precision provided by the stereotactic navigation system allows the surgeon to select the safest possible route to the tumor, make smaller craniotomies, minimize exposure of surrounding normal brain, and make the surgery more elegant. The addition of intraoperative MRI or ultrasound adds a real-time imaging capability to surgery and may further increase the accuracy, particularly after brain shift has occurred. When operating on metastases located in eloquent areas, many surgeons use intraoperative functional mapping of motor or sensory cortex, or perform surgeries under conditions when the patient is awake, with monitoring of speech or motor function throughout resection. The employment of these adjuncts along with refined microsurgical techniques can bring surgical mortality and morbidity to the acceptable levels of 1.6% and 7.7%, respectively.

In each of the prospective studies examining the role of surgery for the management of patients with brain metastases, all patients received WBRT (whole brain radiation therapy)  after surgery. Several investigators subsequently examined whether WBRT is necessary after surgical resection of a brain metastasis. A retrospective review from the Mayo Clinic examined patients who had undergone surgical resection of a single brain metastasis followed by observation or WBRT. The authors found that 85% of the patients in the observation group subsequently experienced brain relapse (defined as local or distant tumor growth) whereas relapse occurred in only 21% of patients in those treated with WBRT. Median survival was longer for the WBRT group as well (21 v 11.5 months).

A randomized trial of observation versus postoperative WBRT in patients with gross total resection of a single brain metastasis showed that recurrence of tumor in the brain was less frequent in the WBRT group than in the observation group (18% v 70%). Furthermore, patients in the WBRT group were less likely to die of neurologic causes than were patients in the observation group (14% v 37%). However, there was no difference in the overall survival (although the study was not powered for survival, only local control) or the length of time that patients remained independent, given that the majority of patients died as a result of their systemic disease. The failure of this and other studies to demonstrate an advantage in survival from the addition of WBRT to surgery can be explained, in part, by the fact that most patients with brain metastases die as a result of systemic progression before one can definitively determine the duration of brain control. In addition, 61% of the patients in the surgery-only group of the study by Patchell  crossed over to receive delayed WBRT, which means that the effect of withholding WBRT altogether on survival was not effectively examined in that study. It is important, therefore, to carefully consider the patient selection criteria used in the studies by Patchell  as well as the relationship between RTOG RPA class and outcome.

In this regard, maximal benefit from aggressive therapy, including postoperative WBRT, is observed in patients who have a KPS score 70, no evidence of leptomeningeal disease, or who are RPA class 1 or 2. Benefit typically is not observed in RPA class 3 patients who have a short survival regardless of choice of therapy. Thus, although there is clear evidence that recurrence within the brain is significantly decreased by WBRT, the impact of this treatment on overall survival will need to be evaluated in the future as the results of systemic cancer improve and survival times increase.

One concern that has been raised regarding the use of WBRT relates to potential long-term effects of this treatment on cognitive function.  Retrospective analyses of relatively small numbers of patients and one large prospective study have raised questions about whether WBRT itself or uncontrolled intracranial metastatic disease plays a greater role in mediating neurocognitive decline.

Potential alternatives to WBRT for maintenance of local control after surgery include radiosurgery, brachytherapy, or intracavitary chemotherapy. Although there are multiple delivery systems available for these modalities, there is insufficient prospective clinical experience available at this time to evaluate their efficacy. Although almost every retrospective study carried out in the last decade called for a multicenter, prospective, randomized trial of stereotactic radiosurgery (SRS) versus surgery for surgically resectable lesions, no such study has been completed. An international randomized phase III trial comparing surgery versus radiosurgery with or without adjuvant WBRT has been continuing enrollment since 1996. Largely because of the noninvasive nature of SRS and the widespread belief that it costs less than surgery, one may think that SRS should eventually replace open surgery, at least for lesions smaller than 3 cm. The lack of unequivocal data to support the superiority of surgery versus SRS only adds to this controversy.

We are aware of five published articles in which investigators directly compared their own use of surgical resection and SRS in the treatment of brain metastases. Bindal   reported on a retrospectively matched series of 75 patients treated with surgery and SRS. The median survival was 7.5 months in the RS group and 16.4 months in the surgical group. The authors concluded that surgery was superior to SRS in the treatment of brain metastases. There is concern, however, that the dosing regimen used in the radiosurgery group resulted in the use of lower prescriptions to the tumor margin than would be considered standard according to the widely accepted RTOG dosing schema.   Hence, underdosing of the radiosurgery group may have led to the inferior results compared with the surgery group.

In contrast, Auchter  retrospectively analyzed a multi-institutional outcome of SRS in 122 highly selected patients with a resectable single brain metastasis. WBRT was performed on all except five patients. The median survival was 56 weeks, comparable with the results of most surgical series. A retrospective study from Muacevic  compared surgery and radiotherapy versus radiosurgery for patients with a single tumor at least 3.5 cm in diameter. The 1-year survival rates (53% v 43%; P = .19), 1-year local control rates (75% v 83%; P = .49), and 1-year neurologic death rates (37% v 39%; P = .8) for the surgical group and SRS group, respectively, were not statistically different. Schoggl   treated 133 patients with either RS (n = 67) or surgery (n = 66) and performed a retrospective, case-control analysis of their results. All patients were treated with WBRT. The median size of the tumors was 7.8 mL for RS and 12.5 mL for surgery, and the median tumor margin dose for RS was 17 Gy. These investigators found no difference in median survival (RS v surgery, 12 v 9 months; P = .19), but that there was a superior local control rate with RS (P < .05). Multivariate analysis showed that the improved local control rate for RS was due to a greater response by radioresistant histologies.

More recently, O’Neill compared these two modalities in treatment of solitary brain metastasis retrospectively and found no significant difference in patient survival. The difference in local tumor control rate, however, was significantly different (100% after SRS and 58% after surgery).

One of the arguments for surgery is the rapid improvement of neurologic symptoms after removal of the mass. However, there have been reports of reduced mass effect and neurologic symptoms after SRS as well. Hoshi reported a series of patients with renal cell carcinoma in which 80% of patients had rapid neurologic improvement after SRS, and complete disappearance of the lesion was seen in 9% of patients. In addition, Sheehan   showed that in patients with non–small-cell lung cancer, 60% of tumors decreased in size, whereas 24% remained stable and 16% enlarged after SRS.

We have summarized the advantages and disadvantages of each treatment modality in Table . Although we attempted to summarize the indications for these two modalities in Table , it is our belief that care should be individualized; not all patients fit precisely into this treatment paradigm. We encourage the strong support of clinical trials for patients with brain metastases.