Fax: (713) 563-3424
Prognostic factors for survival in melanoma patients with brain metastases†
Article first published online: 19 OCT 2010
Copyright © 2010 American Cancer Society
Volume 117, Issue 8, pages 1687–1696, 15 April 2011
How to Cite
Davies, M. A., Liu, P., McIntyre, S., Kim, K. B., Papadopoulos, N., Hwu, W.-J., Hwu, P. and Bedikian, A. (2011), Prognostic factors for survival in melanoma patients with brain metastases. Cancer, 117: 1687–1696. doi: 10.1002/cncr.25634
See editorial and companion articles on pages 1560-3, 1697-703, and 1711-20, this issue.
- Issue published online: 6 APR 2011
- Article first published online: 19 OCT 2010
- Manuscript Accepted: 24 AUG 2010
- Manuscript Revised: 19 AUG 2010
- Manuscript Received: 7 MAY 2010
- brain metastases;
- leptomeningeal disease;
- prognostic factor
One of the most common and deadly complications of melanoma is brain metastases. The outcomes of advanced melanoma patients who developed brain metastases were reviewed to identify significant prognostic factors for overall survival (OS).
An institutional database of advanced melanoma patients enrolled on clinical trials in the Department of Melanoma Medical Oncology from 1986 to 2004 was reviewed and patients who developed brain metastases were identified. Date of diagnosis, patient age, pattern of brain involvement, timing relative to extracranial metastases, prior response to systemic therapy, and treatments given for brain metastases were assessed as potential prognostic factors for OS.
Among 743 melanoma patients enrolled in clinical trials for regional or systemic metastatic disease, 330 (44%) patients developed brain metastases. The median OS after the diagnosis of brain metastases was 4.7 months. Diagnosis before 1996, increased number of parenchymal brain metastases, leptomeningeal involvement, and development of brain metastases after receiving systemic therapy for extracranial metastases were found to be significant prognostic factors for OS. Among patients who received systemic therapy as the initial treatment of brain metastases, patients who previously responded to systemic therapies had longer survival than patients who had not responded.
The era, pattern, and timing of melanoma brain metastases were found to be strongly associated with survival. Previous responsiveness to systemic therapies did not predict better outcomes overall, but it did correlate with improved survival for patients with brain metastases who were treated with systemic therapies. These factors may be used in guiding patient management and for stratifying patients in clinical trials. Cancer 2011. © 2010 American Cancer Society.
Melanoma is the most aggressive form of skin cancer. An estimated 68,720 Americans were diagnosed with melanoma, and 8650 died from the disease, in 2009.1 One of the most common and serious complications of melanoma is the development of central nervous system (CNS) involvement. Melanoma is the third most common diagnosis among patients with brain metastases, after lung and breast cancer.2 Up to 75% of patients with metastatic melanoma develop brain metastases during the course of their disease.3 The development of brain metastases is associated with a very poor prognosis, with a median overall survival (OS) of approximately 4 months reported in several studies.4-6 The significance of this problem is underscored by the finding that up to 54% of all deaths in melanoma patients are attributed to the presence of symptomatic brain metastases.7
Despite these poor outcomes, approximately 5% of patients with melanoma brain metastases are long-term survivors. The identification of factors that can predict outcomes may benefit the clinical management of these patients. In addition, prognostic factors may be useful in the design and interpretation of therapeutic trials for patients with melanoma brain metastases. We reviewed the clinical features and outcomes of patients who were enrolled in clinical trials in the Melanoma Medical Oncology department at The University of Texas M. D. Anderson Cancer Center (UT-MDACC) in Houston, Texas who developed brain metastases during the course of their disease to identify prognostic factors for OS.
MATERIALS AND METHODS
Under an Institutional Review Board-approved protocol, we performed a retrospective chart review of chemotherapy-naive melanoma patients who were enrolled in clinical trials in the Department of Melanoma Medical Oncology at the UT-MDACC from 1986 through 2004. As part of the trial protocols, patients underwent evaluation of the brain by computed tomography or magnetic resonance imaging (MRI) throughout their treatment course and as part of their follow-up. We identified those patients who had brain metastases at any time during their disease course. For these patients, we reviewed primary tumor characteristics, timing and patterns of brain metastases, therapies received, and OS.
For statistical analysis, we used the methods of Kaplan and Meier8 to estimate the median OS. For the analysis of OS, death was the only event. Patients who were lost to follow-up or who were alive at the time of last follow-up were censored at the date of their last follow-up. A Cox proportional hazards regression model was used to test the statistical significance of several potential prognostic factors for OS. This modeling was performed in univariate fashion. Dummy variables were created to represent those variables that have >2 levels. From this model, we estimated the hazard ratio (HR) for each prognostic factor with a 95% confidence interval. Multivariate analysis was performed on those factors that had a univariate P value of <.05.
The records of 743 chemotherapy-naive patients with advanced melanoma who were enrolled on 12 clinical trials were identified and reviewed.9-20 These included 2 trials of neoadjuvant therapy for regional metastases, 1 trial of adjuvant therapy, and 9 trials for metastatic melanoma. We identified 330 (44%) patients who developed brain metastases at some point during the course of their disease. The characteristics of the primary tumors of these patients are summarized in Table 1. The majority of the patients had a cutaneous melanoma, and the most common primary tumor site was the trunk/abdomen (43%). The median Breslow thickness of the 216 cases with available information was 2.4 mm. Although greater than half of the assessable primary tumors measured >2 mm in depth, 18% of patients who developed brain metastases had primary tumors that measured <1 mm deep.
|Primary Tumor Characteristic||No. of Patients/ No. of Patients Evaluable||%|
|Primary tumor site|
|Head or neck||60/330||18.2|
|Breslow thickness, mm|
The median survival from the time of diagnosis of melanoma brain metastases was 4.65 months (Fig. 1A). To determine whether outcomes after the diagnosis of brain metastases had changed significantly over time, we compared survival among patients diagnosed before and after the midpoint of the inclusion dates for this patient cohort (January 1, 1996) (Fig. 1B). Patients diagnosed after this date (n = 123) had slightly longer survival compared with patients diagnosed before that date (n = 207) (median OS, 5.92 months vs 4.14 months; HR, 0.75 [P = .01]) (Table 2). Survival after a diagnosis of brain metastases did not appear to differ significantly between patients diagnosed after the age of 65 years (n = 27) compared with those diagnosed at age ≤65 years (n = 303) (HR, 1.01; P = .97) (Table 2).5
|Variable||No. of Patients (Deaths)||Median Survival, Months||Univariate P||HR||95% CI for HR|
|Brain mets diagnosis date|
|Before 01/01/1996||207 (206)||4.14||—||1.00||—|
|After 01/01/1996||123 (118)||5.92||.01||0.75||0.60–0.94|
|Age at diagnosis of brain mets, y|
|No. of brain mets|
|LMD onlya||11 (11)||1.22||.0002||3.17||1.72–5.85|
|LMD absent||306 (300)||4.87||—||1.00||—|
|LMD present||24 (24)||1.78||.0004||2.12||1.40–3.23|
|Timing vs extracranial mets|
|Before/at extracranial mets||83 (81)||7.93||—||1.00||—|
|After extracranial mets||247 (243)||3.72||<.0001||1.78||1.38–2.30|
|Response to prior systemic therapy|
We examined the pattern of brain involvement in these patients and its impact on patient outcomes. Patients with 1 to 3 parenchymal brain metastases (n = 199) had a median OS of 5.92 months from the time of diagnosis of brain involvement. This was significantly longer that patients with >3 parenchymal brain metastases (n = 118), who had a median OS of 3.52 months (HR, 1.57; P = .0001) (Fig. 2A) (Table 2). Even worse outcomes were observed in patients with leptomeningeal involvement (LMD). Patients with LMD only (n = 11) were found to have a median OS of 1.22 months, which was shorter than that for patients with 1 to 3 (HR, 3.17; P = .0002) or >3 (HR, 2.02; P = .03) parenchymal brain metastases. Overall, patients with LMD (n = 24) had a median OS of 1.78 months from the time of diagnosis of CNS involvement compared with 4.87 months for patients without LMD (n = 306) (HR, 2.12; P = .0004) (Fig. 2B) (Table 2).
The impact of the timing of the development of brain metastases in relation to extracranial metastases was examined (Fig. 2C). Patients who developed brain metastases before the detection of other extracranial metastases (n = 20) had a median OS of 14.26 months from the time of diagnosis of brain involvement, which was longer than that of patients who were diagnosed concurrently with brain and extracranial metastases (n = 63; median OS, 7.0 months [P = .12]). Patients who developed brain metastases after being diagnosed with, and receiving systemic treatment of, extracranial metastases (n = 247) were found to have the worst outcome (median OS, 3.72 months vs patients diagnosed before and/or with synchronous extracranial metastases; HR, 1.78 [P < .0001]) (Table 2).
Patients received a variety of treatments of their brain metastases. The survival of patients who were initially treated with surgical resection (n = 90; median OS, 9.82 months), stereotactic radiosurgery (SRS) (n = 21; median OS, 7.69 months), chemotherapy (n = 55; median OS, 4.64 months), whole brain radiotherapy (WBXRT) (n = 116; median OS, 3.86 months), and supportive care only (n = 35; median OS, 1.22 months) is shown in Figure 3A. It is difficult to retrospectively compare the efficacy of different therapies for melanoma brain metastases because the therapies administered are generally determined by the extent of disease involvement. In general, patients with limited tumor burden in the CNS (usually up to 3 lesions) are often treated with surgical resection or SRS. Patients with more diffuse metastases, or those with poor performance status, are more frequently treated with WBXRT or chemotherapy, or with supportive care only.21 Consistent with this practice, the correlation between the number of brain metastases (1-3 vs >3) and the treatment modality given (craniotomy/SRS vs WBXRT/chemotherapy vs supportive care only) was estimated by the Pearson correlation coefficient, and the P value for this interaction was found to be <.0001. This indicates that there was a significant correlation between the number of brain metastases and the treatment given, and thus prevents meaningful comparison of the relative efficacy of these different treatment modalities in this cohort.
Temozolomide (TMZ) is an alkylating agent that crosses the blood-brain barrier and has produced clinical responses in some patients with melanoma brain metastases.22-24 In this study, patients who received TMZ as the initial therapy for their brain metastases (n = 7) had a median OS of 4.70 months, which did not differ significantly from that of patients receiving any other systemic therapy as their initial treatment (n = 50; median OS, 4.67 months [P = .55]). Patients who received TMZ (n = 24) at any time after the diagnosis of brain metastases demonstrated a trend toward longer survival than patients who received other systemic therapies (n = 134) (median OS, 7.79 months vs 6.92 months; P = .07 by the log-rank test) (Fig. 3B).
It is possible that patients who have chemoresponsive extracranial melanoma metastases will also have responsive brain metastases, and thus will have better survival after the diagnosis of brain involvement. Therefore, we determined the association between survival after the diagnosis of brain involvement and the clinical response to prior systemic therapies. This analysis excluded those patients who were treated in the adjuvant setting. “Responders” were defined as patients who achieved a complete response (CR) or partial response (PR) to prior systemic therapy according to World Health Organization criteria; all other patients were classified as “nonresponders.” Overall, responders to prior systemic therapies (n = 139) had a median OS of 5.36 months after the diagnosis of brain metastases, which was slightly longer but not significantly different from the outcomes for nonresponders (n = 168; median OS, 4.13 months [P = .17]) (Fig. 4A) (Table 2). However, among patients who received systemic therapy as the initial treatment of their brain metastases, the patients who had previously responded to systemic therapy (n = 21) survived significantly longer than those patients who had not previously achieved a clinical response to systemic therapy (n = 31) (median OS, 6.58 months vs 3.81 months; P = .05) (Fig. 4B). Among patients treated initially with WBXRT, there was also trend toward improved survival noted among the responders to prior systemic therapies (n = 52) compared with the nonresponders (n = 56) (median OS, 5.24 months vs 3.24 months; P = .10). It is interesting to note that there was no significant difference noted among the responders and the nonresponders treated initially with systemic therapy in the subsequent use of salvage WBXRT (12 of 21 responders and 14 of 32 nonresponders; P = .34 by the chi-square test). Overall, responders to prior systemic therapy who were treated initially with either chemotherapy or WBXRT had longer survival than the nonresponders (median OS, 5.46 months vs 3.55 months; P = .01) (Fig. 4C). However, the survival for responders treated initially with surgical resection or SRS (n = 47) was longer than that for patients treated with chemotherapy or WBXRT (n = 73) (median OS, 8.62 months vs 5.46 months; P = .02). There was no difference with regard to survival for the responders to prior systemic therapy compared with nonresponders who were treated with surgery or SRS (P = .70) or with supportive care only (P = .87).
Multivariate analysis was performed on variables that were found to be significant on univariate analysis in the overall patient cohort: date of diagnosis of brain metastases, number and pattern of brain metastases, and the timing of brain metastases diagnosis in relation to extracranial metastases. The bimodal comparison of the presence or absence of LMD was not included in this multivariate analysis because of its overlap with analysis of the number/pattern of brain metastases. A diagnosis of brain metastases before January 1, 1996, having >3 parenchymal metastases or LMD only, and the development of brain metastases after receiving systemic therapy for extracranial metastatic disease were factors found to be significantly associated with shorter survival (Table 3). A separate multivariate analysis that included the bimodal absence or presence of LMD and excluded the number of brain metastases identified the presence of LMD (P = .008) as well as the date (P = .004) and the timing (P < .0001) of a brain metastases diagnosis to be prognostic (data not shown).
|Variable||No. of Patients (Deaths)||Median Survival, Months||Multivariate P||HR||95% CI for HR|
|Brain mets diagnosis date|
|Before 01/01/1996||207 (206)||4.14||—||1.00||—|
|After 01/01/1996||123 (118)||5.92||.02||0.75||0.59–0.95|
|No. of brain mets|
|LMD only||11 (11)||1.22||.01||2.29||1.23–4.28|
|Timing vs extracranial mets|
|Before/at extracranial mets||83 (81)||7.93||—||1.00||—|
|After extracranial mets||247 (243)||3.72||<.0001||1.71||1.31–2.22|
The 3 most common sources of brain metastases in patients are lung cancer, breast cancer, and melanoma.21 Because melanoma is much less common than lung and breast cancer, this reflects the high proclivity of melanoma to metastasize to the CNS. Thus, there is a tremendous need to develop more effective therapies for melanoma brain metastases. An improved understanding of the clinical characteristics of these tumors will facilitate research in this area.
In our review of 743 patients with regional or distant metastases who were enrolled in clinical trials at our institution from 1986 to 2004, 44% of the patients with advanced melanoma developed brain involvement during the course of their disease, with a median OS from the time of diagnosis of brain metastases of 4.65 months. This median OS is similar to that presented in other large series that have been reported previously. An analysis of 686 melanoma patients diagnosed between 1985 and 2000 with cerebral metastases at the Sydney Melanoma Unit (Sydney, Australia) reported a median OS of 4.1 months.4 A series of 702 patients with melanoma brain metastases treated at Duke University from 1978 to 1998 reported a median OS of 3.8 months.6 More recently, an analysis of 355 patients with melanoma brain metastases diagnosed from 1991 to 2001 and treated at the Memorial Sloan-Kettering Cancer Center (MSKCC) reported a median OS of 5.2 months.5 Because patients in our cohort who were diagnosed with melanoma brain metastases after 1996 survived longer than those diagnosed before that date, it is possible that the slightly longer median survival in the MSKCC cohort reflects the finding that their analysis focused on more contemporary patients than the other studies.
The increased use of MRI as a screening tool for brain metastases over time may have contributed to this improvement. In addition to possible lead time bias, the earlier diagnosis of brain involvement could possibly lead to an increased frequency of diagnosing patients with lower disease burden, allowing for the more frequent use of locally directed therapies that are associated with better outcomes. In our cohort, we did find a significant difference in the treatments given for brain metastases among patients diagnosed before versus those diagnosed after January 1, 1996 (P = .03 by the chi-square test). Although a similar percentage of patients were treated with chemotherapy or WBXRT (52.5% of patients diagnosed after vs 54.8% of patients diagnosed before January 1, 1996), patients diagnosed after that date were more frequently treated with surgery or SRS (41.7% of patients diagnosed after vs 30.9% of patients diagnosed before January 1, 1996) and less often with supportive care only (5.8% of patients diagnosed after vs 14.2% of patients diagnosed before January 1, 1996). In particular, there was a marked increase in the use of SRS as the initial treatment of brain metastases (2.0% of patients diagnosed before vs 14.1% of patients diagnosed after January 1, 1996), whereas the use of surgery was essentially unchanged (28.9% vs 27.5%, respectively). Although there was no significant difference with regard to patient survival for patients diagnosed after January 1, 1996 versus those diagnosed before that date among those treated with surgical resection (P = .16), chemotherapy (P = .77), WBXRT (P = .51), or supportive care only (P = .48), there was significant improvement in survival noted among patients treated with SRS (median OS, 7.9 months in those diagnosed after January 1, 1996 vs 2.93 months in those diagnosed before; P = .01). It is therefore possible that the improvement in outcomes noted in patients diagnosed after January 1 1996 may reflect the increasing use and efficacy SRS for brain metastases, and not necessarily be attributable only to earlier detection by MRI screening.
In each of the other studies of large cohorts of melanoma patients with brain metastases, the lack of concurrent extracranial metastases was identified as a positive prognostic factor. This factor was also found to be prognostic in our study, although this represents a relatively small population in our cohort. This is likely because the patients being studied herein had sought systemic treatment, usually for distant metastases. Thus, patients who may have had a solitary brain metastasis and then failed to develop additional disease are unlikely to be in this cohort. The MSKCC study also identified patient age >65 years as a negative prognostic factor, but it was not found to be significant in our cohort (Table 2).
Also similar to previous studies, patients with >3 parenchymal metastases were found to have worse outcomes than patients with ≤3 parenchymal metastases in our cohort.5 However, the presence of LMD had a much more dramatic impact on survival, even in comparison with the presence of >3 parenchymal metastases. Similar to cerebral metastases, melanoma is among the solid tumors most frequently associated with LMD. In various reports, melanoma was the associated primary tumor in 17% to 25% of LMD caused by solid tumors.25 In addition, secondary LMD has been identified as a complication of metastatic melanoma in 22% to 46% of patients. There is also an entity known as primary leptomeningeal melanoma, which was first described by Virchow in 1859.26, 27 In our study, we identified 24 patients with LMD. The median survival of these patients was slightly less than 2 months, which was significantly shorter than that of patients with parenchymal involvement only, regardless of the number of metastases. A smaller study on the impact of WBXRT on patient survival in patients with CNS melanoma also identified LMD as a negative prognostic factor.28 In that study of 102 patients, patients with LMD were found to have significantly poorer survival (P = .003) after completing WBXRT compared with patients without LMD. In the MSKCC series, LMD was detected in 11.3% of patients, and they had a median survival of 4 months, which was significantly shorter then the median survival of patients without LMD (6 months) on univariate analysis, but not on multivariate analysis.
Several clinical trials have reported low response rates and poor survival for patients with melanoma brain metastases who are treated with systemic therapies.29-33 Although many of these negative data were not available at the beginning of the study period for our cohort, we did investigate the characteristics of patients who were treated with systemic therapy alone as the initial treatment of brain metastases in our cohort, to better understand what factors might have led to the selection of this therapeutic modality. Comparison of the patients initially treated with chemotherapy with those treated with WBXRT identified no significant difference in the prior responsiveness to systemic therapies among the patients (P = .36 by the chi-square test). Patients treated with WBXRT were more likely to have >3 brain metastases (72.6% vs 30.2%; P < .001). Additional review of patient records identified that the patients treated with chemotherapy were more likely to have active non-CNS metastases that were not considered controlled (92.7% vs 75.2%; P = .007) compared with those patients treated with WBXRT. Thus, overall it appears that systemic therapy was used as the initial therapy for brain metastases in this cohort for patients with lower CNS disease burden and concurrent higher non-CNS disease burden.
Because this cohort of patients was selected based on enrollment in clinical trials for extracranial systemic metastases, we were able to perform a unique assessment of the prognostic value of the responsiveness to previous therapies on survival after the development of brain metastases in a relatively large group of patients. Patients who previously achieved a CR or PR to systemic therapies (“responders”) overall did not have longer survival after the diagnosis of brain metastases compared with those patients who had not achieved a response (“nonresponders”). However, prior response to systemic therapy did predict for longer survival among patients who received systemic therapy as their initial treatment of brain metastases. It is interesting to note that we also observed a trend toward longer survival for the responders compared with the nonresponders among patients initially treated with WBXRT. Although it is therefore possible that the improved survival noted with systemic therapy for brain metastases could be caused by an increased subsequent use of WBXRT, we did not find a significant difference in the use of this salvage therapy between the responders and the nonresponders. In addition, among the patients who did receive salvage WBXRT after initial chemotherapy for brain metastases, there was a longer interval from the initiation of chemotherapy to the initiation of radiotherapy for the responders versus the nonresponders (median, 130 days vs 47 days; P = .01). This possible selective benefit for chemotherapy for brain metastases among patients with documented prior responses to systemic therapies for non-CNS metastases needs to be evaluated in other cohorts of patients. If confirmed, this would suggest preferential benefit for the use of systemic treatments in patients whose tumors previously demonstrated responsiveness to systemic therapies. In addition, combined analysis of patients who were treated with either systemic therapy or WBXRT as the initial treatment of brain metastases demonstrated a significantly longer survival for the responders compared with the nonresponders. These data support a rationale for these somewhat aggressive interventions in patients with documented responses to prior systemic therapies compared with the use of supportive care only. Although comparative analysis of the relative efficacy of different treatment modalities is limited by the retrospective nature of the current study, our data do not specifically support that such modalities should be used in place of surgery or SRS in patients who are considered eligible for such treatments, because responders treated with such local therapies did have longer survival compared with those treated with chemotherapy or WBXRT (P = .02).
In conclusion, this study reported on the factors that were correlated with survival after the development of brain metastases among a large cohort of patients with advanced melanoma who were treated at our institution. Several of the factors that were found to be significant in this cohort have also been identified in previous studies, such as >3 parenchymal metastases, LMD, and the development of brain metastases after prior systemic therapies. This study is distinct from other series with regard to the predominance of patients in this cohort who previously received systemic therapies and the documentation of responsiveness to those treatments. To our knowledge, we have identified for the first time data indicating that patients who previously responded to systemic therapies for extracranial melanoma metastases may derive greater benefit from systemic treatments, or from WBXRT, for brain metastases compared with patients who failed to respond to prior systemic treatments. Our data also indicate that although there has been a statistically significant improvement in the outcomes for patients with melanoma brain metastases over time, the outcomes among these patients remain quite poor. There is a critical need for new, more effective treatments for patients with brain metastases from melanoma, perhaps building on our evolving understanding of the molecular pathogenesis of this disease.34 The results of this study add to our knowledge about outcomes in patients with melanoma brain metastases, and may be useful in future treatments and the design of clinical trials for these patients.
CONFLICT OF INTEREST DISCLOSURES
Supported in part by the Carol Courtney Memorial Fund and the Chiron Corporation.
- 2Metastatic brain tumors. In: Kaye AH, Laws ERJr, eds. Brain Tumors. 2nd ed. Philadelphia: Churchill Livingstone; 2001: 999-1026., , , .
- 3Cutaneous melanoma. In: DeVita VT, Hellman S, Rosenberg SA, eds. Cancer: Principles and Practice of Oncology. 6th ed. Vol 2. Philadelphia: Lippincott Williams and Wilkins; 2001: 2012-2069., , , .
- 141992;11: 343., , , et al. A phase-II study of biochemotherapy using interleukin-2 (IL-2) + interferon alpha-2A (IFN) in combination with cisplatin (C), vinblastine (V) and DTIC (D) in patients with metastatic melanoma. Presented at: 28th Annual Meeting of the American Society of Clinical Oncology, May 17-19, 1992, San Diego, CA.
- 182003;22: 2889., , , , , Phase II study of CTD (cisplatin, paclitaxel, DTIC) in metastatic melanoma (MM). Presented at: 39th American Society of Clinical Oncology Annual Meeting, May 31-June 3, 2003, Chicago, IL.
- 192006;24: 18007., , , , , . Phase II study of “hybrid” biochemotherapy for advanced melanoma. Presented at: 42nd American Society of Clinical Oncology Annual Meeting, June 2-6, 2006, Atlanta, GA.
- 20Long-term clinical results of the combination of cisplatin (C), vinblastine (V), DTIC (D) and interferon-alfa (I) with or without tamoxifen (T) for metastatic melanoma. Presented at: 42nd American Society of Clinical Oncology Annual Meeting. 2006;24: 8039., , , et al.
- 34Analysis of the genome to personalize therapy for melanoma [published online ahead of print August 9, 2010]. Oncogene doi:10.1038.onc.2010.323.,