Integrating personalized ctDNA into clinical surveillance of soft tissue sarcoma: Performance, timing, and histologic insights.

Abstract

11581 Background: Soft tissue sarcoma (STS), particularly larger and biologically aggressive subtypes, carries a high risk of relapse making effective post-operative surveillance essential. CtDNA utility in STS remains elusive. Methods: We conducted a retrospective review of STS who underwent surgical resection and post-operative surveillance with serial personalized ctDNA testing (Signatera) and cross-sectional imaging between April 2023- January 2026. Diagnostic performance was assessed by sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), using imaging-confirmed progression as the reference standard. Subgroup analyses by major histologic subtype conducted. Detection times were referenced to surgery, and lead time was summarized using medians and interquartile ranges (IQR). Paired detection times were compared using the Wilcoxon signed-rank test, with Mann–Whitney U test as a sensitivity analysis. Results: Total 114 patients with ctDNA samples were screened. 101 patients with paired ctDNA ( total 1,005 plasma samples) and imaging surveillance were included. The median age at diagnosis was 53 years, and the cohort was evenly distributed by sex (50.5% male, 49.5% female). Most patients were White (86.1%), 13.9% African American; 58.4% identified as non-Hispanic and 40.6% as Hispanic. Using imaging as the reference standard, 33 patients developed radiographic progression. ctDNA was positive in 27, yielding a sensitivity of 81.8% (95% CI, 65.6%–91.4%). Among 68 patients without progression, 66 remained ctDNA negative, corresponding to a specificity of 97.1% (95% CI, 89.9%–99.2%). The PPV and NPV were 93.1% and 91.7%, respectively. Six patients (18.2%) had radiographic progression despite persistently negative ctDNA, forming a clinically relevant discordant subgroup. Among patients with paired molecular and radiographic progression (n = 27), ctDNA detected progression earlier in 44%, while imaging led in 56%. Overall detection timing did not differ significantly between modalities (median lead time, –0.21 months; IQR, –1.42 to 1.67; p = 0.91). Tumors associated with ctDNA positivity were numerically larger than ctDNA-negative (median, 8.75 cm vs 7.0 cm, p = 0.051). CtDNA detection dynamics varied by histologic subtype, with a nonsignificant numerical trend toward earlier detection in synovial sarcoma. The most common site of metastatic involvement was lung (40%), followed by liver (20%). Conclusions: Personalized ctDNA surveillance showed high specificity and strong diagnostic performance in STS and detected progression earlier than imaging in a subset of patients, though no overall lead-time advantage was observed. Histology-specific differences suggest biologic variability in ctDNA detectability and support prospective, subtype-focused studies to define optimal integration with imaging.