Association of KIT mutations with risk of central nervous system (CNS) metastasis (met) in patients (pts) with mucosal melanoma (MM).

Abstract

9521 Background: The CNS is a frequent site of distant met in pts with cutaneous melanoma (CM). Previous studies have identified clinical, pathological, and molecular risk factors for CNS met in CM pts. However, little is known about the incidence of and risk factors for CNS met in MM pts, who overall have a worse prognosis than CM pts. Methods: We performed an institutionally approved retrospective review of pts diagnosed with clinically localized or regionally metastatic MM at MD Anderson Cancer Center from 1/1/1988 to 12/31/2023. Pts who presented with distant met were excluded. Pt and tumor features (including clinical testing for BRAF, NRAS, and KIT mutations) and distant recurrence events (CNS met; non-CNS met) were assessed. Tumor samples from a subset of MM pts were stained and scored for expression of PTEN (Absent vs Present), PD-L1 (<1% vs ≥1%), and KIT (Above vs Below median H-score) protein by immunohistochemistry (IHC). Time-to-CNS met and Time-to-non-CNS met were computed from the date of initial MM diagnosis (dx) to date of CNS/non-CNS met. Cumulative incidence of distant met events was determined using competing risks (death); pts alive with no met at last follow-up (f/u) were censored. Group differences were evaluated by Gray's test, and associations between measures of interest were determined using proportional sub-distribution hazards regression models. Results: 579 MM pts with clinically localized (73%) or regionally metastatic (27%) disease were included in the analysis. At a median f/u of 34.4 months (range 0 – 525.8), 111 pts (19.2%) had developed CNS met. The cumulative incidence of CNS met at 1, 2, 3 and 5 years was 5%, 10%, 14% and 18%, respectively. For pts with CNS met, median time from MM dx to CNS met was 26.1 months (range, 2.3 – 211.0). Most pts with CNS met presented with brain met only (90%), followed by brain met and LMD (6%), and LMD only (4%). On univariate analysis, KIT mutation (Hazard Ratio [HR] 2.78; 95% Confidence interval [CI] 1.74 – 4.44), p<0.001), mitotic rate 5-9/mm 2 (vs. 0-4/mm 2 ; HR 2.22; 95% CI 1.14 – 4.34, p=0.020), and lymphovascular invasion (HR 1.63; 95% CI 1.03 – 2.56, p=0.036) were associated with increased risk of CNS met. On multivariable analysis, KIT mutation (HR 2.77; 95% CI 1.71 – 4.51, p<0.001) remained significantly associated with increased risk of CNS met. In contrast, KIT mutation predicted a lower risk of non-CNS met in multivariate analysis (HR 0.55; CI 0.34 – 0.87, p=0.011). Among the 87 MM pts for whom IHC was performed on tumor samples, PTEN, KIT, and PD-L1 protein expression were not associated with risk of CNS or non-CNS met. Conclusions: KIT mutation is significantly associated with increased risk of CNS met in patients with clinically localized or regionally met MM. These results highlight distinct CNS met risk factors, and potential surveillance strategies, for MM compared to CM pts.