Leveraging longitudinal clinical annotations and panel sequencing to characterize dynamics of organotropism in metastatic melanoma.

Abstract

9526 Background: Although metastasis plays a major role in the mortality and morbidity of melanoma, patterns of metastasis and drivers of organotropism have not been well characterized clinically. Although previous efforts have been made to examine genomic drivers of organotropism, a longitudinal examination of metastasis dynamics has not been well characterized clinically. Methods: To address these questions, we leveraged a clinico-genomic dataset of patients with metastatic melanoma (Kehl et al. 2021). From this dataset, we assembled a cohort consisting of 26,143 artificial intelligence-annotated imaging reports from 879 patients with metastatic melanoma with an average of 2 years of followup after initial metastatic diagnosis. Annotated metastasis sites included brain, bone, adrenal, liver, lymph node, lung, and mesentery. All patients were evaluated at the Dana-Farber Cancer Institute, and each patient has at least one tumor biopsy sequenced with OncoPanel, a next-generation sequencing panel that identifies mutations in 447 cancer-related genes. Lifetime incidence of metastatic sites was defined as whether a patient ever had a report annotated with that site. Time-to-event analyses on metastatic sites used the acquisition of a metastatic site as the event of interest. For each patient index dates were defined as the date of the first scan with cancer present. We used cumulative incidence models to examine the incidence of each site in our cohort, and both cause-specific hazard models and Fine-Gray modeling to account for competing events and determine the effect of genomic covariates. Results: For lifetime metastasis status, the overall incidence varied by site, with LN being the most common, occurring in ~75% of patients, and mesentery the rarest, occurring in ~12% of patients. For lifetime metastasis status, patients with adrenal metastases were positively enriched for mutations in SETD2 (P < 0.005). Among the significant genes for all sites were many genes encoding epigenetic modifiers. In the time-to-event setting, the order of the cumulative incidence functions for each site matched the lifetime frequency of sites. Through time-to-event analysis we found genes whose mutant status were significantly associated with a change in risk and/or rate of certain metastatic sites. Among other results, we found that having a mutation in SETD2 is significantly positively associated with an increase rate and risk of adrenal metastases (cause-specific HR 2.4, P < 0.005, Fine-Gray HR 2.6, P < 0.005), and NRAS mutations were significantly associated with lower rate and risk of liver metastasis (cause-specific HR 0.59 P < 0.05, Fine-Gray HR 0.58 P < 0.01). Conclusions: By using time-to-event analysis with longitudinal metastatic annotations, we identify potential drivers of organotropism in melanoma patients.