DNA liquid biopsy-based prediction of cancer-associated venous thromboembolism

Authors

Jee J

Brannon AR

Singh R

Derkach A

Fong C

Lee A

Gray L

Pichotta K

Luthra A

Diosdado M

Haque M

Guo J

Hernandez J

Garg K

Wilhelm C

Arcila ME

Pavlakis N

Clarke S

Shah SP

Razavi P

Reis-Filho JS

Ladanyi M

Schultz N

Zwicker J

Berger MF

Li BT

Mantha S

Doi

10.1038/s41591-024-03195-0

PMID: 39147831 · DOI: 10.1038/s41591-024-03195-0 · Journal: Nature Medicine (2024)

TL;DR

This study investigated whether circulating tumor DNA (ctDNA) detected via liquid biopsy can predict cancer-associated venous thromboembolism (VTE). Across three cohorts totaling 6,030 patients with pan-cancer diagnoses (discovery n = 4,141; prospective validation n = 1,426; international generalizability n = 463), ctDNA detection was independently associated with higher VTE rates (HR = 2.49, 95% CI: 1.99–3.11). A random survival forest model trained on liquid biopsy variables outperformed the Khorana score (c-indices 0.74 vs. 0.57 in discovery, 0.73 vs. 0.61 in validation, 0.67 vs. 0.54 in generalizability). In ctDNA-positive patients, anticoagulation was associated with lower VTE rates (adjusted HR = 0.50, 95% CI: 0.30–0.81), whereas ctDNA-negative patients showed no benefit from anticoagulation.

Cohort & data

  • Discovery cohort: 4,141 patients with any cancer type, plasma sequenced with MSK-ACCESS (129 genes) at MSK (June 2019–September 2022). Linked to msk_ctdna_vte_2024.
  • Prospective validation cohort: 1,426 patients with any cancer type, plasma sequenced with MSK-ACCESS at MSK (September 2022–September 2023). Same dataset.
  • International generalizability cohort: 463 patients with advanced NSCLC, plasma sequenced with ctDx Lung (21 genes) at MSK and GenesisCare (Sydney, Australia). Linked to nsclc_ctdx_msk_2022.
  • Cancer types included: NSCLC (34%), breast (13%), pancreatic (10%), melanoma (7%), prostate (5%), bladder (5%), plus 53 other cancer types in the discovery cohort.
  • 61% of discovery cohort patients were ctDNA-positive; 11% developed VTE after plasma draw.

Key findings

  • ctDNA detection was associated with higher VTE rates in the discovery cohort (HR = 2.49, 95% CI: 1.99–3.11, P < 0.001) PMID:39147831.
  • A dose-dependent relationship was observed: patients in the highest VAF quartile had the greatest VTE risk PMID:39147831.
  • The ctDNA-VTE association held across most cancer types, with the notable exceptions of bladder, hepatobiliary, and colorectal cancers PMID:39147831.
  • In multivariate analysis, ctDNA detection (HR = 1.66, 95% CI: 1.30–2.11), cfDNA concentration (HR = 1.61, 95% CI: 1.36–1.92), Khorana score, number of disease organ sites, and chemotherapy receipt were all independently associated with VTE PMID:39147831.
  • The association between ctDNA and VTE was largely independent of tumor genomic content: in a model including tissue-confirmed gene-level alterations, ctDNA detection remained significant while individual gene alterations (including KRAS, STK11, KEAP1) were not PMID:39147831.
  • A random survival forest (LB+ model) using liquid biopsy variables plus cancer type and chemotherapy achieved c-indices of 0.74 (discovery), 0.73 (validation), and 0.67 (generalizability), substantially outperforming Khorana score (0.57, 0.61, 0.54) and the risk assessment model RAM (0.62) PMID:39147831.
  • In ctDNA-positive patients (n = 2,522), anticoagulation was associated with lower VTE rates (adjusted HR = 0.50, 95% CI: 0.30–0.81); in ctDNA-negative patients (n = 1,619), anticoagulation showed no benefit (adjusted HR = 0.89, 95% CI: 0.40–2.0) PMID:39147831.
  • Statin prescription was associated with VTE reduction in ctDNA-positive but not ctDNA-negative patients PMID:39147831.

Genes & alterations

  • TP53: highest unadjusted HR for VTE when detected in plasma (HR = 1.85, 95% CI: 1.53–2.23; n = 1,272 patients with alteration) PMID:39147831.
  • KRAS: adjusted HR = 1.65 (95% CI: 1.30–2.09; n = 465). Known thrombogenic association PMID:39147831.
  • EGFR: adjusted HR = 1.62 (95% CI: 1.24–2.10; n = 339) PMID:39147831.
  • KEAP1: adjusted HR = 2.50 (95% CI: 1.62–3.85; n = 76). Strongest gene-specific association PMID:39147831.
  • STK11: adjusted HR = 1.61 (95% CI: 1.06–2.46; n = 107) PMID:39147831.
  • CTNNB1: adjusted HR = 2.70 (95% CI: 1.74–4.18; n = 65) PMID:39147831.
  • ERBB2: adjusted HR = 1.72 (95% CI: 1.10–2.68; n = 121) PMID:39147831.
  • PIK3CA: adjusted HR = 1.47 (95% CI: 1.01–2.13; n = 254) PMID:39147831.
  • CDKN2A: adjusted HR = 1.84 (95% CI: 1.29–2.63; n = 148) PMID:39147831.
  • However, in multivariate analysis with matched tissue sequencing, individual gene-level alterations were not independently associated with VTE after controlling for ctDNA detection, suggesting the ctDNA-VTE link is largely genomic-content-independent PMID:39147831.

Clinical implications

  • Liquid biopsy-based VTE risk stratification could be provided as a byproduct of standard-of-care ctDNA sequencing already ordered for therapy selection, adding no additional cost or burden to patients or clinicians PMID:39147831.
  • ctDNA detection may identify patients who benefit from prophylactic anticoagulation (adjusted HR = 0.50 in ctDNA-positive patients), while ctDNA-negative patients may be spared unnecessary anticoagulation PMID:39147831.
  • The LB-based model could shift VTE risk assessment from individual providers/hospital systems to assay distributors, who already produce genomic sequencing reports PMID:39147831.
  • Highest-risk patients identified by the model had cumulative VTE incidence exceeding 25%, approximately three times higher than high-risk Khorana score patients PMID:39147831.

Limitations & open questions

  • Observational, non-randomized design: anticoagulation analysis is exploratory and subject to confounders (comorbidities, indication bias). Randomized prospective trials are needed before clinical adoption PMID:39147831.
  • Liquid biopsies are not universally deployed and are primarily used for targeted therapy matching in certain cancer types, limiting immediate generalizability PMID:39147831.
  • The ctDx Lung panel lacks matched white blood cell sequencing, so approximately 11% of mutations may be attributable to clonal hematopoiesis, potentially inflating VTE risk estimates PMID:39147831.
  • Sparse minority representation in the cohorts; predominantly advanced-disease patients at MSK PMID:39147831.
  • ctDNA was not associated with VTE in bladder, hepatobiliary, or colorectal cancers – reasons for cancer-type-specific differences remain unexplained PMID:39147831.
  • Other biomarkers (genetic risk scores, tissue factor-bearing microparticles, proteomics) may provide orthogonal information and could further improve risk models PMID:39147831.

Citations from this paper used in the wiki

  • “ctDNA detection was associated with higher rates of VTE (hazard ratio (HR) = 2.49, 95% confidence interval (CI): 1.99–3.11, P < 0.001)” – Results, ctDNA and VTE.
  • “A machine learning model trained on liquid biopsy data outperformed previous risk scores (discovery, validation and generalizability c-indices 0.74, 0.73 and 0.67, respectively, versus 0.57, 0.61 and 0.54 for the Khorana score)” – Abstract.
  • “anticoagulation was associated with lower VTE rates if ctDNA was detected (n = 2,522, adjusted hazard ratio (HR) = 0.50, 95% confidence interval (CI): 0.30–0.81); ctDNA- patients (n = 1,619) did not benefit from anticoagulation (adjusted HR = 0.89, 95% CI: 0.40–2.0)” – Abstract.
  • “the association between ctDNA and VTE risk appears largely independent of tumor genomics” – Results, multivariate analysis.
  • “cfDNA concentration was the most important feature in predicting VTE in a model with access to all variables” – Results, LB-based models.

This page was processed by paper-compiler on 2026-04-11.