Brain metastases are a frequent complication in advanced melanoma. A 3.6 to 4.1-month median survival has been reported after treatment with whole brain radiotherapy. We performed a retrospective analysis of our institutional experience of multimodality treatment utilizing linear accelerator (Linac)-based stereotactic radiosurgery (SRS).
Forty-four melanoma patients with brain metastases underwent 66 SRS treatments for 156 metastatic foci between 1999 and 2004. Patients were treated with initial SRS if ≤5 brain metastases were present. All patients had Karnofsky Performance Status (KPS) ≥70, but 37 patients had active systemic metastases (Recursive Partition Analysis Class 2). Survival was calculated from the time of diagnosis of brain metastases. Minimum follow-up was 1 year after SRS. The potential role of prognostic factors on survival was evaluated including age, sex, interval from initial diagnosis to brain metastases, surgical resection, addition of whole brain radiotherapy (WBRT), number of initial metastases treated, and number of SRS treatments using Cox univariate analysis.
The median survival of melanoma patients with brain metastases was 11.1 months (95% confidence interval [CI]: 8.2–14.9 months) from diagnosis. One-year and 2-year survivals were 47.7% and 17.7%, respectively. There was no apparent effect of age or sex. Surgery or multiple stereotactic radiotherapy treatments were associated with prolonged survival. Addition of WBRT to maintain control of brain metastases in a subset of patients did not improve survival.
Annually, an estimated 97,000–170,000 cancer patients in the US develop brain metastases.1 Melanoma ranks 4th overall as a cause of brain metastases (10% of patients with brain metastases), after lung and breast cancer, and unknown primary tumors,2 but is 2nd highest in incidence proportion percentage.3 Approximately 10% to 13% of patients presenting with regional disease (AJCC stage III) are at risk for brain metastases,4 and 18% to 46% of stage IV patients will develop central nervous system (CNS) involvement,1, 5 with a prevalence of 55% to 75% at autopsy.5–8 Factors that may associate with development of brain metastases include male gender, mucosal or head and neck primaries, as well as deep or ulcerated lesions.9 Development of brain metastases leads directly to the patient's death in the majority of cases.8, 9 Current management strategies appear unsatisfactory,10, 11 and melanoma patients with brain metastases are usually excluded from participation in clinical trials due to a general perception of an adverse prognosis.
Surgical resection, usually followed by radiotherapy, has been employed to treat brain metastases in selected patients (usually with solitary, superficial lesions).12, 13 Extended survival can be achieved in patients with high performance status or Radiation Therapy Oncology Group (RTOG) Recursive Partition Analysis (RPA) Class I (Karnofsky Performance Status [KPS] ≥70%, primary site controlled, <65 years old, and no evidence of systemic metastases outside the brain).1, 9, 13, 14 Because multiple, rather than solitary, brain metastases are believed to be more common in melanoma,2 the majority of patients with melanoma brain metastases are not considered surgical candidates. In unresectable patients, whole brain radiotherapy (WBRT, generally 3000 cGy in 10 fractions) has been widely employed to palliate brain metastases. In larger series, median survival of patients with melanoma brain metastases treated with WBRT has ranged from 3.6 to 4.8 months.9, 14–16 Fife et al.17 recently published treatment results from 686 patients treated between 1985–2000 at the Sydney Melanoma Unit. These investigators found a median survival of 8.7 to 8.9 months in 205 patients treated with surgery with or without radiotherapy (generally for a solitary metastasis), whereas patients treated with WBRT (236 patients) had a median survival of 3.4 months and patients who received supportive care only (210 patients) had a median survival of only 2.1 months. These results underscore the need for improved treatment algorithms for melanoma patients with brain metastases.
A series of recent publications have suggested high local control rates for brain metastases with stereotactic radiosurgery (SRS), using either linear accelerator (Linac)- or Gamma-Knife-based approaches.18–30 In these studies SRS has generally been limited to patients with 1 to 3 metastases, with rare series accepting patients with larger numbers of lesions.31 We report our encouraging institutional experience, based on expanded eligibility for Linac-based SRS for patients with up to 5 brain metastases, followed by planned systemic therapy.
MATERIALS AND METHODS
All melanoma patients undergoing SRS for brain metastases at the Huntsman Cancer Institute between 1999 and 2004 were identified from records in the Departments of Radiotherapy, Neurosurgery, and Medical Oncology. Retrospective analysis of treatment data was approved by the Institutional Review Board. Minimum potential follow-up of surviving patients was 1 year after treatment with SRS. The following clinical information was obtained: age, sex, date of initial diagnosis of primary melanoma (if known), date of radiographic diagnosis of brain metastases and number of brain metastases, date of SRS, surgical or WBRT, number of SRS treatments, the radiation dose administered by SRS, and apparent cause of death (either systemic, neurologic, or both). The presence or absence of systemic metastases, subsequent systemic therapy, and the eventual cause of death (neurologic, systemic disease progression, or both) were also recorded. Whether patients received treatment with biologic or chemotherapeutic agents after SRS was noted.
Clinical Management Strategy
All stage IIIb or IV melanoma patients (based on 2001 AJCC recommendations32) were screened with a brain magnetic resonance imaging (MRI) or contrast-enhanced computed tomography (CT) scan at the time of initial evaluation at the Huntsman Cancer Institute. Follow-up brain imaging was performed annually for 2 years, or earlier if symptoms warranted. Patients were treated when brain metastases were discovered. SRS was used as the primary treatment modality if there were ≤5 brain metastases. WBRT was usually employed if >5 brain metastases were detected, with stereotactic boost to larger lesions. Patients were closely followed with brain imaging studies (generally gadolinium-enhanced MRI) at least every 2 months, and salvage treatment was employed if there was evidence of radiographic disease progression. This included SRS boost to 1 or more growing lesions or WBRT if ≥5 new lesions were observed. Palliative surgery was considered if there was a dominant symptomatic and surgically accessible lesion, or there was a need for diagnostic tissue.
All patients were considered for early systemic melanoma therapy after SRS if brain imaging was stable or improved at initial follow-up (usually at 1 month). Of 37 patients with systemic metastases, 30 (81%) successfully received systemic treatment with chemotherapy or biologic agents. Most frequently, treatment was with biochemotherapy (20 patients). Regimens utilized during the time of this analysis consisted of CVD biochemotherapy (cisplatin 20 mg/m2 Days 1–4 intravenously [i.v.], vinblastine 1.2 mg/m2 Days 1–4 i.v., DTIC 800 mg/m2 Day 1 i.v., IL-2 9 million IU/m2/day as a 96-hour continuous i.v. infusion and interferon-α 5 million units/m2 Days 1–5, 8, 10 and 12 subcutaneously), or an institutional protocol of T-biochemotherapy (consisting of temozolomide 150 mg/m2 Days 1–4 orally, with IL-2 and interferon-α as described above). A small number of patients were treated with temozolomide alone (200 mg/m2 orally Days 1–5; 8 patients), vaccine (1 patient) or high-dose interleukin-2 (600,000 IU/kg every 8 hours, Days 1–5 and 15–19 i.v.) (1 patient).
SRS Planning and Treatment
Patients were treated with a BrainLAB (Chicago, Ill) m3 micromultileaf collimator attached to a Clinac 21EX or a Novalis SRS unit. High-resolution MRI (2 mm T1-postgadolinium) images were used to define target volumes and risk structures after fusion to CT scans obtained with a stereotactic localizer frame (BrainLAB). BrainSCAN software was used to fuse the images and plan the SRS with dynamic conformal arc techniques. This allowed each lesion to be treated with a single isocenter, with total treatments times under 1 hour for up to 5 metastases. Dose was prescribed to the isodose line, which covered ≥95% of the target volume (range, 74%–96%). Conformity indices were calculated for each lesion. Dose was based on maximal diameter of the metastasis as follows: <2 cm, 22 Gy; 2–3 cm, 18 Gy; ≥3 cm, 15 Gy. Lesions >4 cm were usually not considered for SRS, although rare exceptions were made.
Assessment of Response and Survival
The primary endpoint of the analysis was survival. Survival was defined from the date of CT or MRI scan establishing the diagnosis of brain metastases, as well as date of SRS, until death or last clinical follow-up. Secondary outcomes include cause of death, when applicable (related to systemic metastases vs neurologic complications or both), local control of treated lesions, and overall CNS control. Local failure was defined as an increase in tumor diameter of >25%.
Survival was analyzed by the Kaplan-Meier method33 and potential risk factors were explored using univariate Cox proportional hazard models.34, 35 The effect of the following univariate variables on survival was analyzed: age, interval to brain metastases (years), sex, added WBRT (given as either primary or secondary treatment), initial number of brain metastases treated (1 vs 2 or more), number of radiosurgery sessions (one vs 2 or more), and use of surgical resection. Statistical significance was established at probability levels less than .05. A multivariate Cox regression analysis was performed but not reported. A step-down procedure was used to choose a parsimonious model. All statistical calculations were performed using and Statistica 6.0 software (Tulsa, Okla).
From April 1997 to October 2004, 44 patients with 156 intracranial metastases due to malignant melanoma underwent SRS at the Huntsman Cancer Institute. Tumor volumes ranged from 0.02 to 36.24 cm3. This encompassed all patients with brain metastases due to melanoma treated during this time. There was a male predominance (30 men vs 14 women). The median age was 53 years (range, 26–89). The average interval from initial melanoma presentation to the brain metastases was 3.5 years (median, 2.0 years; range, 0–26 years). Twenty-two patients had a solitary cerebral metastasis and 22 had >1 metastasis (range, 2 to 10). In 7 patients CNS metastases were the only detectable site of disease, whereas 37 patients had brain and systemic metastases. All had a KPS of ≥70%. Five patients met criteria for RPA class I and 39 were RPA class II (2 with CNS-only disease but age >65).
Twenty-six (59%) patients underwent a single SRS treatment, whereas 18 (41%) had 2 to 4 treatment sessions. There was no apparent correlation between the number of brain metastases and the number of stereotactic treatments needed for control (correlation coefficient −0.25), as retreatment generally was employed to treat new lesions. Doses were .5 to 22.56 Gy prescribed to the margin of the tumor volume (median dose, 18 cGy). The most common prescription isodose line was 80% (range, 74%–96%; median 80%).
Local control follow-up was available for 131 of 156 treated brain metastases. Twenty-eight treatments were to a single isocenter and 38 were to multiple isocenters (2–8; median, 3 isocenters). Local control was maintained at last radiographic follow-up for 86 of 131 (65.6%) evaluable lesions. SRS failed to control 45 lesions. The median time to progression of individual treated lesions was 10.2 months. The probability of local control of all treated lesions at 6, 12, and 24 months after SRS was 71.3%, 47.2%, and 36%, respectively (Fig. 1).
Median survival from time of diagnosis of brain metastasis was 11.1 months (95% CI: 8.2–14.9 months) with a range 1.4 to 44 months (Fig. 2). One- and 2-year survivals were 47.7% and 17.7% from diagnosis, respectively. Median survival from first SRS treatment was 9.4 months (95% CI: 7.1–12.5 months) (Fig. 3), due to time elapsed between diagnosis by radiographic scans and radiotherapy/neurosurgery evaluation, treatment planning, scheduling for head-frame application, as well as a small number of patients who underwent initial surgery for diagnosis (with 1–2 lesions) or WBRT for >5 lesion, who subsequently received SRS boost to dominant or growing sites of disease. Thirty patients received SRS as initial therapy for brain metastases. WBRT (before or after SRS) was used in 21 patients, 4 patients received partial brain irradiation but no WBRT, and 19 patients were treated with SRS alone for their brain metastases. Fourteen patients who received WBRT initially presented with a mean of 3.8 metastases, compared with 1.6 initial metastases in 7 patients who received salvage WBRT after CNS progression after SRS. However, overall survival in both groups from diagnosis of brain metastases was not statistically different: 13.1 months in patients with initial WBRT (95% CI: 8–18.3 months) vs 10.5 months (95% CI: 4.8–16.2 months) for patients treated with salvage WBRT. Whereas we suspect that patients with recurrence with >5 metastases after SRS have an adverse prognosis, this was not demonstrable in this small series. Twenty-two patients underwent tumor resection, 14 before and 8 after SRS. The median survival of RPA class I patients was 551 days (95% CI: 329–773 days), whereas RPA class II patients had a median survival of 315 days (95% CI: 211–419 days; P > .05).
A total of 30 of 37 patients who had systemic metastases successfully received systemic therapy after SRS (20 biochemotherapy, 8 temozolomide-based chemotherapy, 1 vaccine-based, and 1 high-dose IL-2). The remaining 7 patients declined systemic treatment due to age, comorbid illness, or individual preference. Toxicity after biochemotherapy appeared similar to patients who did not have brain metastases.
Employing Cox univariate analysis, exploratory evaluation of the effect of age, sex, interval to brain metastases, initial number of metastases, number of stereotactic RT treatments, effect of added WBRT, and surgery was performed (Table 1). There did not appear to be an effect of age younger or older than the median of 53 years, interval from original melanoma diagnosis to brain recurrence greater or less than the median (2 years), number of initial brain metastases, or addition of WBRT. In univariate and multivariate analyses, surgical resection was associated with improved survival, with a median survival of 14.4 months vs 6.5 months in patients not undergoing resection (P = .02). The number of SRS treatments given to control CNS metastases also correlated with survival, with a median survival of 7.4 months with 1 treatment vs 16.2 months with >1 SRS treatment, P = .02.
Table 1. Univariate Analysis of Risk Factors Influencing Survival of Patients With Melanoma Brain Metastases Treated With SRS*
The hazard ratios are estimated using the Cox proportional hazards model. The P values are from the log-rank test.
Gehan Wilcoxon test gives P = .0094 for the effect of surgical resection.
Age <53 (22)
Age ≥53 (22)
Interval to brain mets
<2 y (23)
≥2 y (21)
No resection (23)
Salvage WBRT (21)
No salvage WBRT (23)
Initial no. metastases
1 Mets (22)
≥2 Mets (22)
No. SRS treatments
1 SRS (26)
≥2 SRS (18)
Thirty-eight patients died during the follow-up period (86%). Of these, 6 died of systemic disease progression with controlled CNS metastases and 11 died with concurrent systemic and CNS progression; 21 died predominantly of CNS melanoma. Six (14%) patients currently remain alive.
Patients with advanced melanoma develop brain metastases with such a high frequency that development of appropriate screening and management strategies represents an important clinical challenge. What appears extremely clear is that nontreatment of melanoma brain metastases (‘palliative care’) results in rapid patient decline and death.17, 36 Surgical resection, usually followed by radiotherapy, has been employed to treat brain metastases in selected patients (generally solitary metastases in a superficial location).12, 13 These patients may have extended survival, particularly if they have a high performance status or RTOG RPA class I (KPS >70%, no synchronous systemic metastases).1, 9, 13, 14 Most patients with melanoma brain metastases are not considered surgical candidates. In these patients WBRT (generally 3000 cGy in 10 fractions) has been widely employed to palliate brain metastases, resulting in 3.6 to 4.8-month median survival.9, 14–17, 36 Performance status and RTOG RPA class are important determinants of survival after WBRT.15, 37, 38
We used SRS as the primary treatment for 44 sequential patients with melanoma brain metastases, using a multimodality treatment algorithm. Subsequent systemic therapy was also planned if metastatic disease was not limited to the CNS. In our series half of the patients presented with a solitary brain metastasis. The other half had multiple metastases. Our results indicate that the vast majority of patients with brain metastases from melanoma can be effectively treated using SRS as the mainstay of therapy. SRS demonstrated a high local control rate, with median duration of response in treated lesions lasting 10.2 months and a prolonged median survival (11.1 months, 95% CI: 8.2–14.9 from diagnosis). We found that additional SRS treatments, surgical resection, and WBRT could be employed as an adjunct to initial SRS treatment to maintain control of brain metastases. Exploratory Cox univariate analysis suggested that surgery and multiple SRS treatments appeared to correlate with prolonged survival. Furthermore, 81% of patients with concurrent non-CNS metastases tolerated subsequent systemic treatment without undue toxicity. Although not quantified, quality of life and neurologic function appeared well maintained for protracted periods.
The SRS treatment approach was based on recent publications suggesting high local control rates for brain metastases treated with SRS using either Linac- or Gamma-Knife-based approaches.24, 26–30, 39, 40 In a melanoma-specific patient series, Gonzales-Martinez et al.20 treated 24 patients with 115 lesions with Gamma-Knife (mean radiotherapy dose 16.4 Gy), with a median survival of 5.5 months after radiosurgery. Median tumor volume treated was 4 cm3. Noel et al.19 treated 25 patients with 61 melanoma brain metastases using invasive immobilization and Linac-based radiosurgery (mostly for 1–3 metastases). The median dose was 17 Gy and local control was achieved in 81% at 1 year. Median survival was 8 months, with 29% 1-year survival. Herfarth et al.21 employed Linac stereotactic radiotherapy in 64 patients with 122 melanoma brain metastases, with a median dose of 20 Gy. CNS control was achieved in 81% and median survival was 10.6 months. Radbill et al.23 treated 188 melanoma brain metastases in 51 patients using a Gamma-Knife, achieving 81% local control in the CNS. A 17.7-month survival was achieved in patients with a solitary metastasis, compared with 4.6 months for patients with multiple metastases. RPA class I patients had a median survival of 57 weeks vs 20 weeks for RPA class II or III. The majority of patients had additional brain metastases develop. Selek et al.22 reported the results of Linac-based SRS in 103 patients with 153 brain metastases. The median tumor volume treated was 1.8 cm3, with a median tumor dose of 18 Gy. Local control of treated lesion was 49% and CNS progression-free survival was 15% at 1 year. Overall survival was 25% at 1 year. Koc et al.25 treated 26 melanoma patients with 72 brain metastases using Gamma-Knife. The median survival was 9 months from diagnosis and 1 year survival was 25%.25 Recently, Gaudy-Marqueste et al.30 updated results of Gamma-Knife treatment of 1 to 4 melanoma brain metastases. These investigators treated 221 brain metastases in 106 patients (61.3% were solitary metastases). A high local control rate (83.7) was observed, but median survival was found to be only 5.1 months. A number of conclusions can be drawn from these studies, in conjunction with our own data: SRS or Gamma-Knife treatment appears to produce high control rates in patients with small numbers of brain metastases and good performance status. Some of these studies, including our own, also demonstrated prolonged progression-free and overall survival. In contrast to some published series, single brain metastases and use of additional WBRT did not confer survival benefit.
Because all surviving patients in the current series were followed for a minimum of 1 year after SRS, we can clearly define causes of treatment failure. Six (14%) patients currently remain alive in remission or continue to respond to therapy (only 1 of these initially presented with a CNS-only recurrence, the rest had systemic metastatic disease). Of the 38 patients who died, 6 (14%) died of systemic progression only, 11 (25%) died with concurrent systemic and CNS, and 21 (48%) patients died predominantly of CNS progression. These results indicate that, whereas SRS provides lengthy palliation, maintenance of quality of life and prolongation of survival, further improvement in CNS control is needed, as progression at treated sites and development of new brain metastases contributed to eventual death in ≈73% of patients.
Whether SRS treatment can be used without routine addition of WBRT is currently being evaluated. In a series of renal cancer, melanoma and sarcoma patients treated for 1 to 3 brain metastases using radiosurgery by the Eastern Cooperative Oncology Group, intracranial failure rates were 25.8% and 48.3% at 3 and 6 months, respectively. Failure rates within and outside the treatment volume appeared similar. This translated into a median survival of 8.3 months.39 It is difficult to analyze the impact of SRS in melanoma patients, however, because this was not separately reported. Still, those authors concluded that the intracranial failure rate suggested that routine avoidance of WBRT should be approached judiciously. A recent randomized study by Aoyama et al.41 demonstrated that addition of WBRT to SRS did not significantly improve survival in patients with 1 to 4 brain metastases (mostly lung cancer), with a median survival of 8.0 months with SRS alone, vs 7.5 months with SRS + WBRT. CNS recurrences were seen in 29 of 67 patients after SRS alone, compared with 10 of 65 patients treated with SRS + WBRT. This resulted in frequent addition of salvage WBRT in the SRS-only group. It is clear from our study and that of Aoyama et al that salvage therapy after SRS is feasible without adverse effects on overall survival. The advantage of this approach is avoidance of any unnecessary CNS toxicity from additional radiotherapy (SRS or WBRT) in approximately 50% of patients. The question of whether SRS alone is adequate initial treatment for patients with ≤5 brain metastases with delayed WBRT or SRS salvage vs SRS and immediate WBRT will require a randomized trial to resolve (ideally including neurocognitive and quality of life assessment). It should be noted that patients who require WBRT after failure of SRS for >5 new brain metastases may have less favorable disease, compared with those eligible for SRS retreatment with ≤5 metastases.
In conclusion, we report an encouraging institutional experience treating melanoma brain metastases, based on expanded eligibility for Linac-based SRS for patients with up to 5 brain metastases. Based on our results, we propose the following treatment algorithm (Fig. 4). SRS treatment of melanoma brain metastases followed by systemic treatment provides appears to provide useful palliation and prolonged survival. Due to improvements in linear accelerator design and computerized treatment planning software, even more complex SRS treatment planning is currently possible. Based on the these results, exclusion of appropriately treated patients with melanoma brain metastases from clinical trials does not appear warranted because survival is similar to stage IV melanoma overall. There remains a need to develop more active agents for treatment of metastatic melanoma, with particular emphasis on drugs that penetrate into the CNS to reduce the high incidence of brain metastases.