Predictive factors for the development of brain metastasis in advanced unresectable metastatic melanoma

Background: Melanoma that metastasizes to distant sites is associated with a grave prognosis. The objectives of the study were (1) to identify predictive factors for the development of brain metastases from the time of diagnosis of stage III/IV disease, (2) to identify predictive factors for the development of central nervous system (CNS) metastases from the time of diagnosis of primary melanoma, and (3) to assess whether the incidence of brain metastasis is more frequent in patients who had no tumor response to systemic therapy for stage III/IV disease compared with those who had partial or complete response. Patients and Methods: We collected and retrospectively analyzed information of 740 patients with advanced metastatic melanoma treated at MD Anderson Cancer Center over 15 years. Three hundred and twenty-nine patients had CNS metastases. The characteristics of these patients in terms of median age, sex, primary site, Breslow thickness, stage at first visit, baseline serum parameters, and response to systemic therapy were compared with those of patients who did not develop CNS metastasis. Cox proportional hazards models were used to analyze the cause-specific hazard function for CNS metastasis and deaths without CNS metastasis. Results: We identified that M-stage [stage M1b vs. stage III or M1a, hazard ratio (HR)=2.64; stage M1c vs. stage III or M1a, HR=2.13, P<0.0001] and lactic acid dehydrogenase (LDH) (elevated vs. normal LDH, HR=1.51, P<0.001) at diagnosis of unresectable stage III/IV disease can independently predict the risk of developing CNS metastasis from the time of diagnosis of stage III/IV disease. Older age (HR=1.01, P=0.076), chemoresistance (stable disease+progressive disease vs. complete response+partial response HR=2.91, P<0.0001), low level of albumin (vs. normal HR=2.87, P<0.0001), elevated LDH (vs. normal HR=1.55, P=0.0004), and M-stage (M1c disease vs. stage III or M1a HR=1.89, P<0.0001) can independently predict shorter time to death without CNS metastasis from the diagnosis of stage III/IV disease. The location (head and neck vs. limbs HR=1.56, P=0.028; trunk and abdomen vs. limbs HR=1.45, P=0.029; unknown site vs. limbs HR=8.43, P=0.036) and pathology [Clark level (CL)=3 and/or BR2 to 4 mm vs. CL≤2 and/or BR<2 mm HR=1.60, P=0.037; CL>3 and/or BR> 4 mm vs. CL≤2 and/or BR<2 mm HR=2.03, P=0.001) of the primary melanoma can independently predict CNS metastasis-free interval from the time of diagnosis of primaries. Age (HR=1.012, P=0.034) and pathology of the primary melanoma (CL>3 and/or BR>4 mm vs. CL≤2 and/or BR<2 mm HR=1.54, P=0.024) can independently predict time to death without CNS metastasis from primaries. Conclusion: We identified the predictive factors associated with the development of CNS metastasis in patients with unresectable metastatic melanoma.