Prediction of venous thromboembolism in cancer patients by tissue factor dependent microparticle coagulant activity, biomarkers and a clinical score

Background: Cancer patients receiving chemotherapy in an ambulatory setting are at risk of venous thromboembolism (VTE). Randomized controlled trials have shown that thromboprophylaxis with lowmolecular- weight-heparin reduces the occurrence of VTE in cancer patients but the overall incidence of 4% is generally considered too low to justify thromboprophylaxis in all patients. A logical next step is to focus on tools for selecting cancer patients at the highest risk of VTE, in whom thromboprophylaxis is expected to have a more favorable risk-benefit ratio. Aim: We aimed to determine the performance of tissue factor (TF)- dependent microparticle coagulant activity in predicting VTE, and to compare it to that of the Khorana score, and previously evaluated biomarkers. Methods: In six hospitals blood was obtained from consenting ambulatory patients receiving chemotherapy for stage III or IV cancer. Patients were prospectively followed for 6 months for development of VTE. At inclusion, TF-dependent microparticle coagulant activity was measured in a fibrin generation test (FGT) on fresh plasma in the local laboratories. The outcome measure of this plasma recalcification test was the prolongation of the clotting time in the presence of antibody to factor VII/TF, expressed as percentage of the clotting time in the absence of antibody. The result was considered positive above 13%. Remaining plasma was frozen and stored for later central measurement of D-dimer, P-selectin, pro-thrombin fragment 1 + 2 (F1 + 2) and factor VIII; for these tests positivity cut-offs from the original publications were used. Clinical data were collected for calculation of the Khorana score, which assigns patients to three risk categories. Estimates of sensitivity and positive predictive value (PPV) were calculated, taking into account death as a competing risk, with 95% confidence intervals (CI) based on bootstrap re-sampling. Results: The prospective cohort has recruited 443 patients, with a mean age of 61 years, of which 49% women. In total, 23 patients developed VTE, after a mean time of 2.1 months (5.2%); 77 patients died, after a mean follow-up of 3.3 months (18%). Sensitivity was estimated at 61% (95% CI: 31-84%) for FGT, 84% (61-97) for D-dimer, 85% (56-96) for P-selectin, 44% (61-97) for factor VIII, and 70% (36-89) for F1 + 2. The positive predictive value was 4.7% (1.9-9.3%) for FGT, 6.9% (3.5-12) for D-dimer, 14% (4.8-27) for P-selectin, 11% (4.5-21) for factor VIII and 5.9% (2.4-12) for F1 + 2. A high Khorana score had a sensitivity of 63% (28-83) and PPV of 5.1% (1.6-12). None of the differences were statistically significant. Summary/Conclusions: There are no substantial differences in predictive performance between the studied biomarkers. None of the individual positive predictive values exceeded 15%. In this cohort of patients with advanced cancer, 18% died, which makes adjustment for death as competing risk of utmost importance. In our opinion single predictors will probably not solve the problem and combinations are likely to be helpful