Absolute quantitative proteomics identifies patterns of plasma proteins associated with venous thromboembolism in patients with colorectal cancer
J.T. Buijs a, N. van Es b c, C. Englisch d, R.J.S. Anijs a e, F.T.M. Bosch b f, B.J.M. van Vlijmen a, F.I. Mulder b f, I. Pabinger d, C. Ay d, H.H. Versteeg a 1, Y. Mohammed g h i 1
a) Einthoven Laboratory for Vascular and Regenerative Medicine, Division of Thrombosis and Hemostasis, Department of Internal Medicine, Leiden University Medical Centre, Leiden, the Netherlands
b) Department of Vascular Medicine, Amsterdam UMC location University of Amsterdam, Amsterdam, the Netherlands
c) Amsterdam Cardiovascular Sciences, Pulmonary Hypertension & Thrombosis, Amsterdam, the Netherlands
d) Division of Hematology and Hemostaseology, Department of Medicine, Medical University of Vienna, Vienna, Austria
e) Department of Clinical Epidemiology, Leiden University Medical Center, Leiden, the Netherlands
f) Department of Internal Medicine, Tergooi Hospitals, Hilversum, the Netherlands
g) Proteomics Centre, University of Victoria, Victoria, BC, Canada
h) Gerald Bronfman Department of Oncology, McGill University, Montreal, Canada
i) Center for Proteomics and Metabolomics, Leiden University Medical Center, Leiden, the Netherlands
Abstract
Cancer patients have an eleven-fold increased risk of venous thromboembolism (VTE) compared to the general population. In this hypothesis-generating study we investigated plasma protein levels in colorectal cancer patients with and without thrombosis using targeted absolute quantification of 269 plasma proteins. We included samples from 142 patients with stage III/IV colorectal cancer from MICA – a multinational prospective cohort study, and from 98 patients from CATS – a prospective cohort study with stage III/IV colorectal cancer. The primary outcome was objectively confirmed symptomatic or incidental deep vein thrombosis or pulmonary embolism during a 6-month follow-up period. In MICA, 11 (7.7%) developed VTE; in CATS 6 patients (6.1%) developed VTE within 6 months, and 10 (10.2%) within 2 years.
Six differentially abundant proteins (DAPs) were identified in MICA: APOB100, CD5L, IGHG1, IGHM, PRG4, and TF. A model using these six proteins achieved a c-statistic of 0.687 (95% CI: 0.658–0.717) by internal cross-validation with 100 repeats and random 80% training sets. The optimism-corrected c-statistics was 0.675 (95% CI: 0.648–0.701). Using any subset of the DAPs yielded a c-statistics >0.67, with the best model reaching 0.768 (95% CI: 0.746–0.790). This outperformed the Khorana, modified-Vienna, PROTECHT, and CONKO models. The six DAPs had also predictive value in CATS with c-statistics at 0.70 (95% CI:0.671–0.728) for the 6-month, and 0.73 (95% CI:0.702–0.728) for 2-year follow-up. Two proteins showed inverted patterns between cohorts, likely due to chemotherapy. Our findings call for further investigation into the proteins identified, warranting further validation in larger, standardized studies.