In patients with active cancer who develop venous thromboembolism (Ca-VTE), major bleeding emerges as the most serious nonfatal complication during anticoagulant (AC) therapy. The range of possible risk factors for bleeding is broad, and predicting the bleeding risk in such scenarios proves challenging [1]. Clinical trials involving patients with VTE show that the case-fatality rate for major bleeding events fluctuates between 8 and 11%, yet evidence from real-world cohorts points to significantly higher fatality rates, ranging from 18 to 20% [2-3].
Among the previously published existing VTE bleeding risk scores, only the “CAT BLEED Score” was derived from a cohort specifically comprising patients with Ca-VTE [4]. The other scores were developed based on cohorts that either did not include or only partially included patients with both VTE and cancer, likely affecting their performance due to reduced study power and inherent limitations [5].
Most bleeding risk scores have used clinically relevant non-major bleeding (CRNMB) as an outcome.
The outcome in the new B-CAT study was clinically significant bleeding (either ISTH major bleeding or bleeding leading to hospitalisation and not all CRNMB. That is because although CRNMB not leading to hospitalisation are very frequent they can often be relatively minor and therefore are not used to influence treatment decisions.
A recent study utilizing real-world data from the UK Clinical Practice Research Datalink (CPRD) aimed to develop a relevant risk prediction model for Ca-VTE patients who are at risk of bleeding during AC therapy [6].
Methodological Framework: Developing the B-CAT Scoring System
This retrospective, observational study analyzed a cohort of 15,749 patients from 2008 to 2020 who experienced venous thromboembolism (VTE) during active cancer (Ca-VTE), AC treatment, and at risk for a bleeding event. The average age of the cohort was 66.3 years. The study defined Ca-VTE as any VTE event occurring within six months of cancer diagnosis, treatment or upon documentation of disease progression or metastases. The types of deep vein thrombosis (DVT) considered included thrombosis in the legs’ deep veins, calf, iliac veins, and vena cava, excluding atypical DVTs such as those in the cerebral, upper limb, and splanchnic veins [6].
Cancer diagnoses were obtained from hospital and GP records, noting that the most prevalent cancers in females were breast (n=2,906), ovarian (n=5,62), and uterine (n=379), whereas in males it was prostate cancer (n=1,639). The leading non-sex-specific cancers included colorectal and other lower gastrointestinal tract (n=2,605), lung (n=1,538), hematologic (n=1,418), and upper gastrointestinal tract cancers (n=994). At the time of Ca-VTE, 7,052 patients (44.8%) had metastatic disease [6].
Regarding AC treatment, nearly 60% of the AC-treated Ca-VTE cohort received parenteral AC, 18.4% were given vitamin K antagonists (VKA), and 12.9% were treated with direct oral anticoagulants (DOACs) [6].
The outcome of interest was significant bleeding in outpatients, characterized as either major bleeding (including fatal bleeding, critical site bleeding, or cases of posthemorrhagic anemia) or clinically relevant non-major bleeding requiring hospitalization (CRNMB-H). Baseline covariates, encompassing both known and suspected risk factors for these bleeding events, were collected at the outset of the observational period. Time-varying covariates such as CRNMB not requiring hospitalisation or trauma were defined based on information recorded after VTE diagnosis and before the end of the observational period [6].
On the first day of recorded Ca-VTE, the cohort’s baseline characteristics were summarized. This included categorical variables presented as numbers (proportions) and continuous variables as mean (standard deviation), with data presented for the entire cohort and segmented by VTE type (DVT and PE) and gender [6].
Multivariate Fine and Gray regression models were applied to predict significant bleeding event risk. These models, which consider the competing risk of mortality and adjust for baseline covariates, helped derive sub-distribution hazard ratios (SHRs) for all examined predictors of significant bleeding. [7] A scoring system based on these SHRs was created using an additive scale to predict significant bleeding events [6].
The predictive accuracy of the developed scoring system was validated through its discrimination capability, which distinguishes between patients who will and will not experience a bleeding event, confirmed by C-statistic estimation. Calibration was also evaluated by grouping accumulated score points into “low,” “medium,” and “high” risk categories, with subsequent calculation of bleeding event incidence rates for each group [6].
Findings from the B-CAT Score Evaluation
Over 4,914 person-years, the study observed 537 significant bleeding events within the first six months of anticoagulant (AC) therapy, split into 161 major bleeds and 376 CRNMB-H. Major bleedings were primarily due to intracranial (37.9%) and gastrointestinal (32.3%) events, while the most frequent CRNMB-H occurrences were hematuria (26.1%), followed by upper gastrointestinal tract bleeding (18.4%) and epistaxis (11.7%), with gastrointestinal bleeds being the most prevalent significant bleeding type, making up 34.8% of CRNMB-H cases [6].
Colorectal and other lower gastrointestinal tract cancers served as the reference for cancer type in this cohort. Seventeen factors were identified as independent predictors of significant bleeding; each awarded one score point. These encompassed various cancers such as bladder, CNS, cervix, kidney, malignant melanoma, prostate, and upper gastrointestinal, plus one point for metastatic cancer, and were extended to include minor surgery and trauma, any history of major bleeding, CRNMB within the last two years, CRNMB-NH post-initial Ca-VTE, and comorbidities [6].
At 6 months of follow-up the bleeding incidence rates per 100 person-years were delineated by risk groups: 5.12 for low risk (0 to 1 point), 18.96 for medium risk (2 to 3 points), and 56.54 for high risk (4 or more points), translating to about 1 in 3.6 high-risk patients experiencing a significant bleed. Major bleeding rates were 1.13, 5.97, and 23.76, and CRNMB-H rates were 3.99, 12.99, and 32.78 per 100 person-years for low, medium, and high risk, respectively, indicating a significant rate difference exceeding eightfold between the lowest and highest risk categories. Low-risk patients accounted for 30.4% of the bleeding events, medium risk for 56.8%, and high risk for 12.8% [6].
Conclusions
This study introduced the B-CAT score, developed from 17 independent risk factors, to accurately quantify the risk of significant bleeding in Ca-VTE patients, occurring in 3.4% of the cohort. Over a 6-month observation period, aligning with the typical duration for anticoagulation due to its associated high risk of bleeding, the study assessed the effectiveness of this scoring system. [6] The continuation, dosage adjustment, or cessation of anticoagulation beyond this timeframe is generally influenced by weighing the risks of VTE recurrence against the risks of bleeding [8].
Showing greater precision than current VTE bleeding risk models, such as the CAT BLEED score, the B-CAT score enhances predictive accuracy with superior C statistics (0.70 (0.65-0.75) for clinically significant bleeding risks, and 0.76 (0.68-0.84) for major bleeding, thereby improving anticoagulation management in Ca-VTE patients [4].
The time varying nature of this model means that it is a dynamic model that can be applied throughout the patient journey incorporating changes in health status or the development of new risk factors.
However, more studies are needed to delineate bleeding risks after the initial 6 months of anticoagulation in Ca-VTE patients [6].
References
- Kearon C, Akl EA, Ornelas J, et al. Antithrombotic therapy for VTE disease: CHEST guideline and expert panel report. Chest, (2016);149(02):315–352
- Carrier M, Le Gal G,Wells PS, RodgerMA. Systematic review: case fatality rates of recurrent venous thromboembolism and major bleeding events among patients treated for venous thromboembolism. Ann Intern Med, (2010);152(09):578–589
- Lecumberri R, Alfonso A, Jiménez D, et al; RIETE investigators. Dynamics of case-fatalilty rates of recurrent thromboembolism and major bleeding in patients treated for venous thromboembolism. Thromb Haemost, (2013);110(04):834–843
- de Winter MA, Dorresteijn JAN, Ageno W, et al. Estimating bleeding risk in patients with cancer-associated thrombosis: evaluation of existing risk scores and development of a new risk score. Thromb Haemost, (2022);122(05):818–829
- den Exter PL, Woller SC, Robert-Ebadi H, et al. Management of bleeding risk in patients who receive anticoagulant therapy for venous thromboembolism: communication from the ISTH SSC Subcommittee on Predictive and Diagnostic Variables in Thrombotic Disease. J Thromb Haemost, (2022);20(08):1910–1919
- Cohen AT, Wallenhorst C, Choudhuri S, Nassar A, Pollok KG, Martinez C. A Novel Risk Prediction Score for Clinically Significant Bleeding in Patients Anticoagulated for Venous Thromboembolism with Active Cancer. Thromb Haemost, (2023)
- Satagopan JM, Ben-Porat L, Berwick M, Robson M, Kutler D, Auerbach AD. A note on competing risks in survival data analysis. Br J Cancer, (2004);91(07):1229–1235
- Lyman GH, Carrier M, Ay C, et al. American Society of Hematology 2021 guidelines for management of venous thromboembolism: prevention and treatment in patients with cancer. Blood Adv, (2021);5(04):927–974