Circulating tumor DNA (ctDNA) in patients (pts) with metastatic uveal melanoma (UM) treated with protein kinase C inhibitor (PKCi).

e22054 Background: UM is the most common primary intraocular malignancy. About 50% of pts with UM will develop metastatic disease and currently there is no effective treatment in metastatic UM. Nearly 95% of UM harbour mutually exclusive activating mutations in GNAQ, GNA11, CYSTLR2 and PLCb4. We sought to evaluate ctDNA in metastatic UM pts receiving experimental early phase clinical trial of LXS196, a PKCi, using digital droplet PCR (ddPCR) and targeted ion torrent next generation sequencing (NGS). Methods: 17 pts with metastatic UM were identified from a single institution. Pt characteristics including mutation status, disease volume using sum of product of disease (SPOD), best response and clinical benefit defined as partial response (PR) or stable disease (SD) > 6 months were assessed. Plasma samples at baseline and early on treatment (EOT) (14 – 30 days on treatment) were analysed for ctDNA using mutation specific ddPCR. A custom made NGS panel covering 95% of UM mutations was used on plasma ctDNA samples at baseline and on treatment. The allele frequency (AF) of activating mutations identified using ion torrent NGS analysis were then validated using ddPCR. Results: Using ddPCR, 16/17 pts had a +ve ctDNA at baseline with baseline ctDNA copies correlating with LDH (n = 17, p = < 0.001, Spearman’s rank r = 0.8015) and SPOD (n = 17, p = 0.005, r = 0.6642). 16/17 pts had paired samples at baseline and EOT and only 4/16 pts were undetectable at EOT. These 4 pts were GNA11 Q209L positive, had below median SPOD (median 5986, range 200 – 16782), and low numbers of liver metastases (median 9, range 1 – 49). Of these 4 pts, one had PR and three had SD (two with SD > 6 months) as best response. A further 8 pts had +ve ctDNA at baseline and showed a reduction in EOT (reduction range 46 – 99%). Of these 8 pts, one had PR, four had SD (two with SD > 6 months) and three had progressive disease (PD) as best response. The remaining 4 pts showed increasing ctDNA from baseline to EOT and all had SD/PD. Using ROC analysis, EOT ctDNA predicted clinical benefit to PKCi (AUC 0.84, [95% confidence interval, 0.65-1.0, p = 0.026]). AF of ion torrent NGS correlated significantly with ddPCR AF (n = 30, p = < 0.001, r = 0.968). Ion torrent NGS was able to detect additional mutations implicated in UM prognosis including SF3B1 mutations in 4 pts. Conclusions: Baseline ctDNA correlates with baseline LDH level and disease volume. EOT ctDNA predicted clinical benefit to PKCi. The ctDNA AF derived from ddPCR and NGS was comparable and targeted ion torrent NGS was useful in detecting driver as well as additional mutations in UM.