Prognostic Factors for Survival in Melanoma Patients with Brain Metastases

Introduction

Melanoma is the most aggressive form of the common skin cancers and is the cause of over 75% of the deaths attributed to skin cancer each year. Worldwide, its incidence is rising, particularly in younger individuals. An estimated 70,000 new cases of melanoma will be diagnosed in the United States alone in 2014 ( http://seer.cancer.gov 2014). Melanoma generally metastasizes regionally first to lymph nodes. While patients with regional lymphatic spread are generally treated with, and can be cured by, surgery, such patients are at high risk for distant metastases. The distant metastatic spread of melanoma cells can affect virtually any site in the body, including the brain parenchyma and/or the leptomeninges. Among common cancers, melanoma is the most likely to metastasize to the brain and is the third-most frequent cause of brain metastases, after breast cancer and lung cancer ( ). The incidence of brain involvement for patients with metastatic melanoma is up to 43% in the clinical setting and up to 75% in autopsy series ( ). While a small subset (<10%) of melanoma patients with central nervous system (CNS) involvement survive greater than 5 years, the median overall survival (OS) for melanoma patients with CNS metastases is only 4–6 months, which is worse than for any other metastatic site in this disease ( ). CNS involvement is the cause of death in up to 54% of patients who die from melanoma ( ).

An improved understanding of both the molecular changes that occur in this disease and the factors that regulate the antitumor immune response is now revolutionizing the treatment of patients with metastatic melanoma. Since 2011 three different targeted therapies (vemurafenib, dabrafenib, trametinib) have demonstrated superior outcomes compared to chemotherapy in randomized phase III trials, leading to their regulatory approval ( ). In addition, combined treatment with dabrafenib and trametinib was approved based on significant increases in clinical response rates and progression-free survival (PFS) versus single-agent therapy ( ). Notably, each of those agents was approved for use specifically in patients with activating BRAF mutations, which are present in approximately 50% of patients with cutaneous melanoma ( ). In parallel to these developments in targeted therapy, immunotherapy with ipilimumab, an antibody that blocks the inhibitory CTLA4 receptor on the surface of T cells, was approved for the treatment of patients with metastatic melanoma in 2011. While ipilimumab has a relatively low clinical response rate (<10%), up to 25% of patients appear to achieve durable (i.e., >3 years) disease control and survival with this treatment ( ). More recently, several experimental therapies that target another molecule on the surface of T cells, particularly programmed death-1 (PD-1), have demonstrated clinical responses in ≥50% of metastatic melanoma patients in early-phase clinical trials, with many responses appearing to be long-lasting (>2 years).

These advances and promising results have given rise to a new era and optimism in the management of this highly aggressive disease. However, to date few of these agents have been evaluated systemically in melanoma patients with brain metastases, and the CNS is increasingly a common first site of treatment failure. In this changing landscape of treatment options, understanding the prognostic factors associated with the development of brain metastases, and with survival after their development, will facilitate the appropriate design and interpretation of new clinical trials, and with clinical decision-making in individual patients. Thus, this review will summarize the existing literature on prognostic factors for brain metastasis development and survival in patients with melanoma.

Factors Associated with CNS Involvement by Melanoma

The cumulative risk of developing brain metastasis within 5 years of a melanoma diagnosis is about 7% ( ). Brain metastases are the first site of distant metastasis in 10–20% of melanoma patients ( ). As summarized in Table 19.1 , a number of retrospective studies have investigated factors associated with the risk of or time to the development of brain metastasis in patients with melanoma. These studies were identified using PubMed search terms “melanoma, brain, metastases, prognosis.” Most of these studies were single center and conducted between 1976 and 2012. The reported incidences of brain metastasis varied between 8% and 46% with most studies reporting an incidence of approximately 10%. No single factor was significantly associated with the risk of brain metastasis universally in these studies, but some associations were identified repeatedly. For example, several studies demonstrated an association for male sex with an increased risk of brain metastasis. Age was only found to be an associated factor in two studies.

The factor associated with the greatest risk for the development of brain metastasis across the studies is primary melanoma location in the head and neck region ( P < 0.001 to P = 0.05). A number of other primary tumor characteristics known to predict survival in melanoma patients that are incorporated in the 7th edition of the American Joint Committee on Cancer staging system for melanoma (Breslow thickness, ulceration, and mitotic rate, ) were also analyzed for an association with brain metastasis. Primary tumors with a higher Breslow thickness had an increased likelihood ( P < 0.0001 to P = 0.029) of metastatic spread to the brain in six of the eight studies that analyzed this factor. Furthermore, the presence of primary tumor ulceration was also associated with a higher risk of brain metastasis in four of seven studies ( P < 0.0001–0.007). Finally, a higher mitotic index was significantly associated with risk of brain metastasis in all three studies evaluating that factor ( P < 0.0001 to P = 0.02), and the nodular melanoma subtype showed a similar association in three of the listed studies ( P < 0.0001 to P = 0.0012). A number of other factors were evaluated in single studies only that did not find an association with the development if brain metastases. These factors included the presence of extracapsular spread ( P = 0.47), nodal stage and nodal region ( P = 0.72), number of involved nodes ( P = 0.36), size of largest resected lymph node ( P = 0.08), sentinel lymph node biopsy ( P = 0.36), interval from initial diagnosis to stage IV diagnosis ( P = not reported), race ( P = not reported), and presence of pigmentation ( P = not reported).

Although melanoma patients have historically been risk-stratified by demographics and primary tumor characteristics, recent advances in therapy are increasingly leading to assessments of tumor molecular features as part of routine clinical care. Early data in this area suggests that some of these features may be associated with CNS metastasis. For example, recurrent hot-spot mutations in BRAF (40–45% of cutaneous melanomas) and NRAS (15–20%) have been assessed in multiple large cohorts of melanoma patients. In a study of 677 patients with stage IV melanoma, 47% of patients had BRAF mutations, and 20% had NRAS mutations. Analysis of the sites of involvement at the initial diagnosis of stage IV disease showed that the patients with BRAF or NRAS mutations were more likely to have brain metastases (24% of those with mutant BRAF and 23% of those with mutant NRAS ) than patients in whom both genes were wild type (12%; P = 0.008) ( ). More recently, an analysis of a cohort of stage IIIB/C melanoma patients failed to identify a significant association between BRAF or NRAS mutation status and time to brain metastasis ( ). However, a significant association was observed between loss of expression of the tumor suppressor PTEN and brain metastasis. Loss-of-function mutations and deletions of PTEN have been detected in 10–30% of melanomas, commonly in tumors with concurrent BRAF mutations but generally not with NRAS mutations. PTEN is a negative regulator of the PI3K-AKT pathway, and loss of PTEN has been shown to result in constitutive activation of the pathway in multiple tumor types, including melanoma. Loss of PTEN expression in the stage III cohort was significantly associated with shorter time to brain metastasis in patients with concurrent activating BRAF mutations ( P = 0.03) but not in patients with wild-type BRAF and NRAS ( ). Another study has recently also reported the molecular analysis of melanoma brain metastases and extracranial metastases, including a subset of patients with multiple metastases available. Although overall patterns of copy number variations, mRNA expression, and protein expression were similar between within-patient paired samples of brain metastases and extracranial metastases, the brain metastases demonstrated higher expression levels of several activation-specific protein markers in the PI3K/AKT pathway than the extracranial metastases ( ). Increased activation of the PI3K-AKT pathway was also found in an immunohistochemical analysis of melanoma patients who had undergone synchronous resection of brain and non-CNS metastases ( ). Together, these findings suggest that the PI3K-AKT pathway should be further investigated as a potential therapeutic target for melanoma brain metastases.

Murine studies have demonstrated that transfection of constitutively activated signal transducer and activator of transcription 3 (STAT3) enhanced brain metastasis of melanoma, whereas transfection with dominant-negative STAT3 suppressed brain metastasis ( ). These highly metastatic melanoma cell lines over expressed matrix metalloproteinases-2 (MMP-2) and blockade of activated STAT3 by expression of dominant-negative STAT3 suppressed MMP-2 expression, prevented invasion, inhibited tumor growth, and prevented metastasis in vivo indicating that there is a central role for STAT3 signaling in the process of metastasis. Further confirmation of the relevance of STAT3 in the process of metastasis comes from tissue microarray studies that have demonstrated higher levels of expression of activated STAT3 in human brain melanoma metastasis specimens compared to primary tumors ( ). Specifically, in 51 primary and 48 brain metastasis specimens obtained from patients with melanoma, only 43% of the former had moderate to strong p-STAT3-positive immunohistochemical staining, whereas 81% of the latter had moderate to strong p-STAT3-positive staining supporting the contention that p-STAT3 is involved in the process of metastasis. We have retrospectively identified 299 patients with stage IV melanoma and assembled a tissue microarray of systemic non-CNS metastasis specimens. Using immunohistochemical analysis to measure the percentage of cells with p-STAT3 expression and Kaplan–Meier survival estimates to analyze the association of p-STAT3 expression with median survival time, time to first CNS metastasis, and development of CNS metastasis, we did not find an association with the development of CNS metastasis; however, p-STAT3 expression was a negative prognostic marker for OS ( ). A limitation of this study was that the primary tumor that was sampled for p-STAT3 expression may have not reflected the biology of the primary tumor at the time of CNS metastasis or that a minority clone may have been sufficient to initiate metastasis.

In summary, a number of factors have been identified as conferring increased risk of development of CNS metastasis in melanoma. Factors that have been identified in multiple studies include male gender; primary tumor in the head and neck region; and primary tumor Breslow thickness, ulceration, and mitotic index. Clinicians should have a heightened awareness of these factors when discussing prognosis with their patients, and potentially when deciding on appropriate clinical follow-up and monitoring. Moving forward, there will be a need to perform integrated analyses of these factors with emerging molecular markers in this disease.

Prognostic Factors for OS in Melanoma Patients with Brain Metastases

Patients with melanoma brain metastases have historically had very poor outcomes. As summarized in Table 19.2 , the median survival after the diagnosis of brain metastases across multiple studies has been 4–6 months ( ). While patient outcomes are poor overall, some patients do achieve durable long-term survival. Thus, multiple analyses have been undertaken to identify factors that correlate with survival in melanoma patients with CNS involvement.

Most studies of outcomes in melanoma patients with CNS involvement have examined the prognostic significance of the number of brain metastases present. While cutoffs used in these analyses have varied, in general the presence of multiple metastases has predicted shorter OS. As shown in Table 19.2 , of the 13 studies that evaluated the number of brain metastases, all but three reported a strong association between single brain metastases and longer OS ( P < 0.0001 to P = 0.03 for single metastases and P = 0.07 to P = 0.61 for more than one). While it is possible that the presence of multiple brain metastases represents a different biology than single brain metastasis, it is important to note that there is also a strong association between the number of brain metastases and the treatments that patients receive. In multiple series, long-term survival has been observed in some patients who underwent surgical resection as a definitive treatment for their brain metastases ( ). Long-term survival has also been observed for patients treated with stereotactic radiosurgery (SRS). Historically these treatments have been limited to patients with a small burden of CNS disease, including generally <3 brain metastases. Patients with more extensive involvement have generally been treated with whole-brain radiation therapy (WBRT), chemotherapy, or supportive care. Comparison of patients treated with surgery or SRS to those treated with WBRT, chemotherapy, or supportive care have shown that that patients treated with surgery or SRS have better OS ( ). Thus, the effect of the number of metastases on the treatments given may contribute to the significant association with OS observed. However, the possibility of a distinct biology does exist.

The prognostic significance of concurrent extracranial disease has been evaluated repeatedly in melanoma patients with CNS involvement ( Table 19.2 ). Of the 12 studies that evaluated the prognostic significance of concurrent extracranial disease in patients with melanoma brain metastases, seven found that the presence of other extracranial metastases was associated with significantly shorter OS (in one trial: hazard ratio, 1.558; P < 0.0001). The detrimental effect of uncontrolled extracranial disease on OS in patients with melanoma brain metastases was initially described in 1978 ( ). The best survival was observed in patients with CNS only disease, without any evidence of other visceral involvement. In one of the largest cohorts of patients examined to date, patients with isolated brain metastasis (n = 380) had a median OS of 135.2 days, while patients with coexisting lung metastasis (n = 191) had a median survival of only 93.8 days ( P = 0.0014), and patients with more than one additional site of visceral metastasis (n = 77) had a median survival of only 39 days ( P = 0.0036) ( ). Of note, the same group described a small group of 17 patients with metastatic melanoma who survived for more than 3 years, 15 of whom had metastatic disease limited to the CNS alone suggesting that CNS metastasis in of itself is not a sole determinant for survival.

Other factors have also correlated with survival. Good performance status and lower lactate dehydrogenase level at the time of diagnosis of melanoma brain metastases was associated with significantly longer OS in two studies ( ). Conversely, patients with neurological symptoms had worse survival ( ). The presence of hemorrhage prior to treatment with SRS was a predictor of neurological death and neurological symptoms ( ). The presence of neurological symptoms can be due to large or numerous tumors or to a tumor in an unfavorable location in the brain, all of which were shown to be negatively associated with survival.

Some studies have analyzed the association between primary tumor features and prognosis. In general these analyses have shown mixed results ( Table 19.2 ). While the presence of ulceration of the primary tumor was associated with shorter survival after the development of brain metastases in one study, two other reports did not find a significant association ( ). Similarly, primary tumor location, histologic subtype, and mitotic index were associated with shorter survival after brain metastases in some studies but not in others ( ). On a molecular level, genes associated with the T-cell receptor pathway were associated with prolonged OS in patients with melanoma with brain metastases, whereas genes associated with hypoxia and oxidative stress were associated with poor OS in these patients, underlining the importance of immune response in the brain environment ( ).

Overall, these results support that the number of brain metastases and the extracranial disease status are key prognostic factors for OS in melanoma patients with CNS involvement. As discussed previously, these factors have also been significantly associated with the treatment modalities used in patients ( ). As the treatment landscape for melanoma evolves, it will be important to reevaluate the prognostic significance of these factors. Of note, our single-center experience in patients with brain metastases demonstrated that the OS of these patients has improved over time (median OS 4.14 months for patients diagnosed before 1996 vs 5.92 months diagnosed after January 1, 1996), even prior to the testing and approval of the new, effective systemic therapies ( ). The development of new focal radiation techniques, such as SRS, may lessen the prognostic impact of the number of brain metastases on survival. Similarly, new therapies with unprecedented rates of disease control also provide hope for the improvement of outcomes in brain metastasis patients with concurrent extracranial disease. Improvements may also be achieved if such systemic therapies also have significant antitumor activity in the brain.