AUTHORS
Debolina Pramanik, Internal Medicine Resident, Hong Li, PhD2, Ravi Trivedi3, Kayvon Sanjasaz4, Chase J Wehrle5, Federico Aucejo6, Keith McCrae, MD7, Alok Khorana, MD8
AFFILIATIONS
1 Cleveland Clinic Foundation, Cleveland, Ohio, United States
2 Cleveland Clinic Foundation, Cleveland, United States
3 Cleveland Clinic Foundation, Cleveland, United States
4 Cleveland Clinic Foundation, Cleveland, United States
5 Cleveland Clinic Foundation, Cleveland, United States
6 Cleveland Clinic Foundation, Cleveland, United States
7 Cleveland Clinic Foundation, Cleveland, Ohio, United States
8 Cleveland Clinic Foundation, Cleveland, Ohio, United States
ABSTRACT
INTRODUCTION
Cancer-associated thrombosis (CAT), including venous thromboembolism (VTE) and arterial thromboembolism (ATE), is a major cause of morbidity and mortality in patients with cancer. Emerging data suggest that presence of cell-free DNA (cfDNA), now commonly used in oncology practice for prognostication and treatment selection, may be associated with risk of CAT. We evaluated the role of cfDNA and risk of CAT in gastrointestinal (GI) malignancies.
METHODS
We conducted a cohort study in patients ≥18 years with GI malignancies who prospectively underwent cfDNA testing with Guardant 360 or Guardant Reveal between 2/2017- 4/2025 at Cleveland Clinic, with a longitudinal follow up of at least 6 months. cfDNA was defined as the total extracellular DNA from all cellular sources, while somatic alterations (SA) were cancer-associated somatic variants. CAT includes deep venous thrombosis (DVT), pulmonary embolism (PE), or ATE (stroke or myocardial infarction). The date of the first ctDNA collection was considered baseline. The primary objective was to determine if cfDNA parameters – maxVAF [highest variant allele fraction (VAF) of all cfDNA mutations], maxSA (highest VAF of SA), total number of cfDNA alterations (totMUT) and SAs (numSA) - were associated with risk of CAT within 6 months. Secondary objectives included identifying specific genetic alterations and risk factors for CAT; comparing predictive ability of Khorana score (KS) and cfDNA parameters for CAT, and association of cfDNA with overall survival (OS). Cumulative incidences were calculated with death as competing risk. The association of cfDNA with CAT within 6 months from cfDNA testing was evaluated using survival analysis. cfDNA parameters were categorized in 4 quartiles. Prediction was explored using area under curve (AUC) at 6 months. OS was calculated using Kaplan-Meier estimates. Patients were followed until the earliest of 1st CAT event, death, last follow-up or study cut-off date (4/2026).
RESULTS
The study cohort comprised 696 individuals; median age at cancer diagnosis was 62 years, with 425 (61.1%) males, 555 (80%) White, 62 (8.9%) Black, 646 (92.8%) non-Hispanic. Of these, 286 (41.1%) had colorectal cancer, 169 (24.3%) had hepatocellular carcinoma, 150 (21.4%) had cholangiocarcinoma, 91 (13.1%) had pancreatic cancer. At baseline, median [Interquartile range (IQR)] of cfDNA was 1.3 (0.3-8.8) for maxVAF, 0.5 (0-6.1) for maxSA, 4 (2-7) for totMUT and 2 (0-4) for numSA.
The 6-month cumulative incidence of CAT was 7.1% (95% CI 5.3–9.2); this included 48 CAT events (21 DVT, 16 PE, 7 DVT+PE, 2 ATE and 2 ATE+DVT). Increasing quartile levels of all 4 cfDNA parameters were associated with increased 6-month incidence of CAT (p for trend < 0.05 for all). The cumulative incidences for the highest quartiles were 10.5% for maxVAF, 10.2% for maxSA, 10.9% for totMUT and 10.0% for numSA. Hazard ratio (HR, 95% CI) of CAT for top vs lowest quartile was 2.07 (0.90, 4.79) for maxVAF, 2.96 (1.17, 7.52) for maxSA, 2.4 (1.04, 5.46) for totMUT and 2.53 (1.06, 6.0) for numSA. The trend remained significant for totMUT (p 0.04) and numSA (p 0.02) after adjusting for age, disease duration, KS and metastatic status at baseline. The most common cfDNA mutations were ATM (16.7%), TERT (11.6%), and ARID1A (10.6%). Only PIK3CA mutations (8.5%) were significantly associated with increased risk of CAT (HR, 95% CI: 2.30, 1.08, 4.93). Metastatic disease (HR 1.84, 95% CI: 1.036-3.265, p 0.04) was associated with increased risk of CAT. Median OS (95% CI) was 30.2 months (26.0, 36.9). The highest quartiles for all cfDNA parameters were associated with significantly worsened OS.
In prediction analysis, cfDNA parameters had similar c-index as KS (c-index 0.61-0.65). When cfDNA parameters were combined with KS, c-index increased (c-index 0.68-0.70).
CONCLUSION
cfDNA parameters – including highest VAF of all cfDNA mutations and SA, total number of cfDNA alterations and SAs - were significantly associated with increased 6-month risk of CAT and decreased survival. PIK3CA mutations were also significantly associated with CAT. However, cfDNA itself was not sufficiently predictive; adding cfDNA parameters to KS appeared to increase the predictive ability. Large prospective studies are needed to validate the utility of adding cfDNA parameters to clinical risk scores to optimize individualized CAT risk assessment.
DISCLOSURES
Debolina Pramanik, Internal Medicine Resident: No relevant conflicts of interest to declare.