Clinicopathologic and molecular characterization of <i>ARID1A</i> -mutant solid tumors using paired tissue and liquid biopsy sequencing.

e15054 Background: ARID1A is a commonly altered chromatin-remodeling gene across solid tumors, with roles in DNA damage repair, tumor immunogenicity and therapeutic response. The clinical significance of ARID1A alterations detected in liquid biopsy, particularly at low variant allele frequency (VAF), remains uncertain. We performed a large clinicopathologic and molecular analysis to define tissue–plasma concordance and clinically relevant VAF thresholds. Methods: Solid tumor cases with tumor tissue and/or liquid biopsy next-generation sequencing (NGS) performed on the MDA MAPP platform at MD Anderson Cancer Center (2019–2025) were analyzed. Tissue–plasma concordance was evaluated in paired samples. Concordance, associations, and overall survival were assessed using standard statistical methods. Results: Among 17,092 patients with NGS data, 1,033 (6.0%) harbored ARID1A mutations in liquid biopsies across multiple tumor types, including colorectal (n = 218, 21.1%), lung carcinomas (n = 228, 22.1%), breast invasive ductal (n = 107, 10.4%), and cholangiocarcinoma (n = 88, 8.1%). Across ARID1A variants, 48.9% were pathogenic/likely pathogenic and 51.1% were variants of unknown significance/benign. ARID1A -mutant tumors showed high genomic complexity, with co-alterations in TP53 (81.4%), APC (41.3%), PIK3CA (33.5%), and KRAS (27.6%). Across 1,303 plasma-detected ARID1A variants, circulating tumor DNA (ctDNA) VAF was distributed as follows: 40.1% &lt; 1%, 30.1% 1–5%, 19.9% 5–20%, and 10.0% &gt; 20%. A total of 303 patients had paired tissue/plasma sequencing. At the patient level, tissue-plasma concordance was complete in 66.7%, partial in 22.4%, and absent in 10.9% of patients. At the variant level, tissue–plasma concordance increased stepwise with higher VAF: 55.7% at &lt; 1% (44/79), 84.7% at 1–5% (61/72), 89.7% at 5–20% (52/58), and 86% at &gt; 20% (26/30). Using a plasma VAF cutoff of 1%, variants with VAF ≥1% had significantly higher tissue–plasma concordance than those with VAF &lt; 1% (85.6% vs 55.7%; OR 4.74, 95% CI 2.60–8.65; p = 8.96×10⁻⁷ ). Concordance by tumor type was compared between variants with VAF &gt; 1% and &lt; 1%, with highest concordance observed in colorectal adenocarcinoma (85.0% vs 59.3%; p = 0.01 ), lung adenocarcinoma (90.6% vs 61.1%; p = 0.02 ), and cholangiocarcinoma (95.0% vs 66.7%; p = 0.15 ). In exploratory analyses, tumors with ARID1A mutations had inferior overall survival compared with ARID1A –wild-type tumors (log-rank p = 0.0002 ). Conclusions: ARID1A alterations detected in liquid biopsy are common but variably concordant with tumor sequencing. Plasma ctDNA VAF is a determinant of reliability, with a ≥1% threshold identifying ARID1A variants with high tissue concordance, particularly in colorectal cancer and lung adenocarcinoma. Integrating VAF, pathogenicity, and tumor context is essential for clinical interpretation of ARID1A liquid biopsy results.