RISK ASSESSMENT MODELS (RAM) OF VENOUS THROMBOEMBOLISM (VTE) IN PATIENTS WITH BLOOD CANCERS: A SYSTEMATIC REVIEW

Blood cancer patients face a relative risk of VTE 28fold higher than the general population. Nevertheless, they also face a significantly higher risk of bleeding than overall cancer patients, thus prevention of VTE is not systematically implemented in this population. A more detailed estimation of VTE risk might support personalized primary prevention decisions. We therefore aimed at reviewing the VTE risk score that have been validated in blood cancer patients. According to GRADE-18 guidelines, we conducted at Nov 2020 a systematic search of literature databases (Cochrane Library, EMBASE, PubMed/MEDLINE) including the studies reported at meetings since Jan 2010. The search was limited to the following blood cancers: multiple myeloma (MM), lymphoma, acute leukemia (AL). Six-month VTE rates for risk categories, sensitivity, specificity, AUC or C-statistics and predictive values of the retrieved scores were recorded. Moreover, PROBASt score was applied for quality appriasal. We retrieved 82 records which resulted in 9 VTE RAMs: 3 ones were devloped in patients with MM, 5 ones in patients with lymphoma and only one was retrieved for AL. Sensitivity of lymphoma RAMs ranged from <50% to 93% and specificity from 54% to 90%. AUC ranged from 0.579 to 0.783. The negative predictive vale of the 5 RAMs was consistently higher than 95% except for the Khorana score (88%). The positive predictive value, however, was very low, ranging from 20% to 35%. Also C-statistics was higher for RAMs specifically developed in lymphoma patients (0.76-0.88) rather than Khorana score (0.60). The expected rate of VTE in high-risk patients ranged from 16% to 25% in lymphoma-specific RAMs, while it was lower than 15% for Khorana score. The expected rate of VTE in low-risk patients was lower than 3.8% in lymphoma specific RAMs. As a result, prophylaxis decisions based on lymphoma specific RAMw achieved a number need to treat lower than 10, namely less than 10 high-VTE-risk lymphoma patients need to receive primary prophylaxis in order to avert 1 VTE. Throly score reported the highest quality. In MM patients, four RAMS were validated: sensitivity ranged from 40% to 85% and specificity from 43% to 72%. C-statistics of the RAMs was poor, ranging from 0.51 to 0.66. Only the IMPEDE-VTE score allowed to discriminate risk categories differing by >10% VTE risk at 6 months. In conclusion, disease-specific VTE RAMs outperform general RAMs, however they still have suboptimal positive predictive values. Specific bleeding RAMs are also awaited.