Circulating Tumor DNA Predicts Venous Thromboembolism in Minority Patients with Cancers

Introduction: Despite rapid advances in liquid biopsy for circulating tumor DNA (ctDNA), its prognostic value for venous thromboembolism (VTE) in cancer patients is underexplored, particularly in underserved and minority populations. This study investigated the association between ctDNA and VTE risk in a diverse cohort of cancer patients.

Methods: We analyzed data from 1,038 cancer patients who underwent ctDNA measurement using the Tempus xF liquid biopsy assay at a large safety-net hospital system in the US. The association between ctDNA and VTE was examined after adjusting for cancer type, stage, treatment, and time from initial diagnosis using Fine and Gray models. We assessed the discrimination of genetic, clinical, and combined models using the area under the time-dependent receiver operating characteristic curve (AUC).

Results: The study cohort was predominantly Hispanic (50.1%) and non-Hispanic Black (32.0%), with prostate (26.4%) and lung (19.5%) cancers being most common. Most patients (68.3%) had metastatic disease. The median follow-up was 11.4 months. Incident VTE occurred in 142 patients, with a cumulative incidence of 9.0% at 6 months and 11.3% at 12 months.

A dose-dependent effect was observed between the number of pathogenic ctDNA mutations and VTE incidence, which remained after stratification by cancer subtype. Higher numbers of ctDNA alterations and higher maximum variant allele fraction were observed in VTE patients compared to non-VTE patients. Administration of chemotherapy, cancer type, and VTE-risk cancer group were also significantly associated with VTE risk.

The presence of pathogenic ctDNA mutations was independently associated with VTE risk after adjusting for clinical variables. The number of pathogenic ctDNA mutations was predictive of future VTE risk across tumor types (adjusted subdistribution hazard ratio 2.75, 1.94, and 1.38 for ≥3, 2, and 1 pathogenic mutation, respectively, compared to none, p<0.0001). The association was primarily driven by mutations in KRAS, PTEN, CDKN2A, NF1, and EGFR genes. The cumulative incidence of VTE at 6 months ranged from 5.5% for patients with no ctDNA mutations to 18.7% for those with ≥3 pathogenic mutations. Compared to the clinical model (AUC 0.71, 95% CI 0.64-0.76), the combined clinical and ctDNA model showed minor improvement in discrimination for VTE prediction (AUC 0.74, 95% CI 0.67-0.80). The number of pathogenic ctDNA mutations was also associated with overall survival in a dose-dependent manner. Compared to patients with no pathogenic mutations, the hazard ratio for mortality ranged from 2.67 for 1 mutation to 5.69 for ≥3 mutations.

Conclusions: ctDNA testing may serve as an adjunctive tool to clinical risk assessment models in cancer patients to improve personalized VTE risk stratification and management. The number of pathogenic ctDNA mutations was independently associated with increased VTE risk across cancer types, even after adjusting for clinical factors. While ctDNA analysis provided additive value to clinical risk models, it did not outperform them, highlighting the need for multimodal prediction approaches. These findings require external validation, particularly using different ctDNA panels. Further research is needed to elucidate the biological mechanisms underlying the association between specific mutations and VTE risk in cancer patients.