Comparison among three predictive models for cancer-associated thromboembolism in a hispanic population

In this study we aimed to evaluate and compare the overall performance of the Khorana, PROTECHT, and CONKO scores as predictive scores for the occurrence of venous thromboembolism (VTE) among ambulatory Hispanic patients with solid tumors. We included all outpatients with newly diagnosed solid tumors receiving systemic chemotherapy in Hospital San Juan Dios, San José, Costa Rica, from January to December 2021. For each patient the Khorana, PROTECHT, and CONKO scores were calculated at the beginning of treatment. The sixth-month cumulative incidence of VTE was estimated using the Fine & Gray competing risk model. The receiver operating characteristic (ROC) curve was used to assess the performance of each predictive tool through the analysis of the c-statistic, sensitivity, and specificity. A total of 708 patients were included in the research. After a median follow-up of 8.13 months, the cumulative VTE incidence at six months was 4.45% (95%CI: 3.25-6.91%) for the overall population. At the conventional positivity threshold of 3 points, these scores classified from 17.7 to 32.5% of all patients as high-risk for VTE. Patients belonging to the high-risk category of the Khorana, PROTECHT, and CONKO scores had significantly higher risk of VTE in comparison to low-risk patients (Khorana score: Hazard Ratio (HR): 2.66; 95%CI:1.20-5.89; p = 0.042; PROTECHT score: HR: 3.44; 95%CI:1.63-7.21; p = 0.001; CONKO score HR: 3.68; 95%CI:1.72-7.85; p = 0.001). The c-statistic of the ROC curve was: 0.62 (95%CI: 0.52-0.72), 0.62 (95%CI: 0.52-0.73), and 0.65 (95%CI: 0.56-0.76) for the Khorana, PROTECHT, and CONKO scores, respectively; with similar sensitivity (range: 67-70%) and specificity (range: 52-62%) among them. For Hispanic patients with solid tumors the Khorana, PROTECHT, and CONKO scores accurately categorize their risk of VTE. However, the overall discriminatory performance of these models remains poor (c-statistic from 0.62 to 0.65) for predicting all patients at risk for thromboembolic events.