Prediction of Venous Thromboembolism (VTE) in Patients with Pancreatic Cancer Using Clinical Data, Biomarkers, and VTE Risk Models

Abstract

Abstract 3398 INTRODUCTION:VTE is common in patients with cancer and causes significant morbidity and mortality. Clinical risk models and biomarkers including C-reactive protein (CRP), soluble P-selectin (sPsel), and D-dimer have been used to predict VTE in diverse groups of cancer patients at varying risk for VTE. The applicability of these findings to specific high risk subtypes of cancer has not been established. Therefore, we sought to identify the value of clinical factors, plasma biomarkers, and risk models in predicting VTE in patients with pancreatic cancer, a malignancy with a high predilection for VTE. METHODS:Patients seen at the University of Michigan Comprehensive Cancer Center (UMCCC) and previously consented and enrolled in a prospective cohort study were eligible. Inclusion criteria are diagnosis of pancreatic adenocarcinoma, evaluation at UMCCC, no VTE within a month prior to cancer diagnosis, and documentation in the Electronic Medical Record (EMR) at least every 6 months until death. Primary objective was to identify factors predictive of VTE. Secondary objectives were to develop a VTE predictive model, assess the utility of published VTE risk models, and evaluate factors associated with overall survival (OS). Demographics, clinical data, and VTE (deep vein thrombosis [DVT], portal vein thrombosis [PVT], or pulmonary embolism [PE]) rate were obtained from the EMR. ELISAs were performed for CRP, D-dimer, Mac-2 binding protein, soluble E-selectin (sEsel), and sPsel using banked plasma specimens drawn at diagnosis. A retrospective cohort study was performed including univariate and multivariate regression analysis. The utility of predictive models by Khorana, et al (Blood, 2008. 111:4902–4907), which includes cancer site, body mass index (BMI), hemoglobin (Hb), platelet (plt) count, and white blood cell count, and the expanded model by the Vienna Cancer and Thrombosis Study (CATS) (Blood, 2010. 116:5377–5382), which additionally includes sPsel and D-dimer, were assessed. RESULTS:Between 2005 and 2011, 89 patients were eligible for analysis. Median follow-up was 268 (18–2433) days. Twenty (22%) cases had a VTE; 10 (50%) DVT, 2 (10%) PE, 4 (20%) PVT, and 4 (20%) multiple VTEs. Mean (SD) age was 63.4 (8.9) in cases and 65.3 (11.2) in controls. Women accounted for 55% of cases and 48% of controls.Higher BMI (median 28.8 [21.2–44.7] in cases vs. 25.4 [16.4–43.3] in controls, p=0.03) and lower plt count (median 241 [145–323] in cases vs. 289 [97–648] in controls, p=0.001) were associated with VTE on univariate analysis. On multivariate regression analysis, lower plt count (β −0.01, SE 0.004) and lower Hb (β −0.43, SE 0.20) were predictive of VTE after adjusting for BMI, tumor location, and treatment with surgery, chemotherapy or radiation (AUC 0.78). None of the biomarkers were significantly associated with VTE on univariate analysis, although there was a trend with D-dimer (p=0.09).The Khorana score was determined in 85 patients; 48 were intermediate (2 points) and 37 high risk (≥3 points) with VTE rates of 20.8% and 24.3%, respectively (p=0.70). The AUC of this model was 0.63. The risk score from CATS was calculated for 84 patients; 54 were intermediate (2 or 3 points), 17 high (4 points), and 13 highest risk (≥5 points). VTE incidence was not different among these groups and the AUC was 0.65.Factors associated with poor OS on univariate analysis were: age (per 10-year increment) (HR [95% confidence interval], p-value) (1.35 [1.07–1.71], 0.013), chronic kidney disease (5.67 [2.62–12.25], <0.0001), use of anticoagulation (3.14 [1.33–7.41], 0.009), stage III/IV vs. I/II pancreas cancer (2.05 [1.27–3.32], 0.003), and INR (1.65 [1.04–2.63], 0.035); elevated Hb (0.87 [0.76–0.99], 0.041) and sEsel (0.46 [0.29–0.72], 0.0007) were protective. CONCLUSIONS:Pancreatic cancer patients with higher BMI, lower plt count, and lower Hb were more likely to develop VTE. Other clinical variables and biomarkers did not add additional predictive information. Elevated sEsel, important for neutrophil trafficking to sites of inflammation, was found to be protective on survival analysis. The risk models developed by Khorana, et al and CATS in a diverse group of patients with cancer were not able to further differentiate VTE risk among this already high risk group. Additional work is needed to determine which patients with pancreatic cancer are at highest risk for VTE and who may benefit most from thromboprophylaxis. Disclosures:No relevant conflicts of interest to declare.