1089 - Genomic Signature for Leptomeningeal Progression in Brain Metastasis Patients Using Liquid Biopsy-Acquired Comprehensive Genomic Profiling

C. M. Lanier¹, A. R. Choi¹, R. DAgostino Jr², S. E. Glynn¹, M. Abdulhaleem³, M. Smith⁴, Y. Wang⁴, J. Ruiz⁵, T. Lycan⁵, W. J. Petty⁵, C. K. Cramer¹, S. B. Tatter⁶, A. Laxton⁶, J. White⁶, C. T. Whitlow⁷, F. Xing⁸, and M. D. Chan¹; ¹Department of Radiation Oncology, Wake Forest University School of Medicine, Winston-Salem, NC, ²Department of Biostatistics and Data Science, Wake Forest University School of Medicine, Winston -Salem, NC, ³Department of Hospital Medicine, Wake Forest University School of Medicine, Winston-Salem, NC, ⁴Department of Molecular and Cellular Bioscience, Wake Forest University School of Medicine, Winston-Salem, NC, ⁵Department of Internal Medicine, Section of Hematology and Oncology, Wake Forest University School of Medicine, Winston-Salem, NC, ⁶Department of Neurosurgery, Wake Forest University School of Medicine, Winston-Salem, NC, ⁷Department of Diagnostic Radiology, Wake Forest University School of Medicine, Winston-Salem, NC, ⁸Department of Cancer Biology, Wake Forest University School of Medicine, Winston-Salem, NC

Purpose/Objective(s): Biomarkers for leptomeningeal disease (LD) are not well elucidated in the literature. Few studies have shown a correlation between genomic data and LD. Materials/

Methods: Patients with brain metastases were identified in our departmental database. The electronic medical record was used to identify patients with liquid-biopsy-based comprehensive genomic profiling (Guardant Health). LD was defined as patients with clinical, cytologic (lumbar puncture) or radiographic evidence of LD. LD risk scores were developed based on the number of genetic mutations they have that are associated with time to LD. Cox proportional hazards regression analyses were used to assess the association between mutations and progression to LD. Hazard ratios were calculated to quantify the risk of developing LD in those with specific genetic mutations compared to those without. A score of +1 was assigned for every mutation present associated with LD with a threshold of p<0.1 to be entered into the model. Patients were then placed into one of three risk groups (0 genes present, 1 gene present, 2+genes present). Product limit survival estimates were developed to provide information on the probability of survival without LD.
Results: Two-hundred and four patients met study criteria and were included in the analysis. Twelve gene mutations were associated (p < 0.1) with developing LD (APC (p = 0.014), ARID1A (p=0.04), CDH1 (p=0.09), DDR2 (p=0.04), EGFR (p = 0.06), MPL (p=0.017), MSH2 (p=0.016), NTRK3 (p = 0.005), PMS2 (p = 0.036) PTPN11 (p=0.09), RAF1 (p=0.033)and SMO (p = 0.021). When risk groups were constructed, 102 patients had none of the twelve genes present, 82 patients had one of the twelve genes present, and 20 patients had at least 2 genes present. We found that 4/102 (3.9%) of the patients with 0 genes, 10/82 (12.2%) with 1 gene, and 10/20 (50%) with 2 or more genes developed LD (Chi Square-34.2, p<0.0001). The Cox proportional hazards regression model indicated that the 3-level genetic risk score was highly associated with time to LD (Wald Chi-Square=18.8, p<0.0001). This model also showed that patients with 2+ genes had a hazard ratio (HR) for time to LD of 12.7 (p < 0.0001) when compared to patients with 0 genes, and patients with 1 gene had a HR of 4.6 (p=0.01) for time to LD when compared to patients with 0 genes. Using the product limit survival analysis model, we determined the 6-month, 1- and 2-year cumulative incidence of LD estimates were 0%, 3%, and 3% for patients with no genes present, 2%, 4%, and 18% for patients with one gene present and 11%, 11%, and 18% for patients with 2 or more genes present, respectively.
Conclusion: Genomic profiling appears to provide a meaningful LD risk score from non-invasive liquid biopsies. Those with higher risk scores were more likely to have LD in their lifetime with certain mutations predisposing a patient at an earlier time point. Validation of this signature could lead to a biomarker with the potential to influence clinical trial enrollment and surveillance mechanisms in this population.