Developing a machine learning model for bleeding prediction in patients with cancer-associated thrombosis receiving anticoagulation therapy.

BACKGROUND: Only 1 conventional score is available for assessing bleeding risk in patients with cancer-associated thrombosis (CAT): the CAT-BLEED score. OBJECTIVES: Our aim was to develop a machine learning-based risk assessment model for predicting bleeding in CAT and to evaluate its predictive performance in comparison to that of the CAT-BLEED score. METHODS: We collected 488 attributes (clinical data, biochemistry, and International Classification of Diseases, 10th Revision, diagnosis) in 1080 unique patients with CAT. We compared CAT-BLEED score, Ridge and Lasso logistic regression, random forest, and Extreme Gradient Boosting (XGBoost) algorithms for predicting major bleeding or clinically relevant nonmajor bleeding occurring 1 to 90 days, 1 to 365 days, and 90 to 455 days after venous thromboembolism (VTE). RESULTS: The predictive performances of Lasso logistic regression, random forest, and XGBoost were higher than that of the CAT-BLEED score in the prediction of bleeding occurring 1 to 90 days and 1 to 365 days after VTE. For predicting major bleeding or clinically relevant nonmajor bleeding 1 to 90 days after VTE, the CAT-BLEED score achieved a mean area under the receiver operating characteristic curve (AUROC) of 0.48 ± 0.13, while Lasso logistic regression and XGBoost both achieved AUROCs of 0.64 ± 0.12. For predicting bleeding 1 to 365 days after VTE, the CAT-BLEED score achieved a mean AUROC of 0.47 ± 0.08, while Lasso logistic regression and XGBoost achieved AUROCs of 0.64 ± 0.08 and 0.59 ± 0.08, respectively. CONCLUSION: This is the first machine learning-based risk model for bleeding prediction in patients with CAT receiving anticoagulation therapy. Its predictive performance was higher than that of the conventional CAT-BLEED score. With further development, this novel algorithm might enable clinicians to perform personalized anticoagulation strategies with improved clinical outcomes.

Does the clot burden as assessed by the Mean Bilateral Proximal Extension of the Clot score reflect mortality and adverse outcome after pulmonary embolism?

Background: Rapid diagnosis and risk stratification are important to reduce the risk of adverse clinical events and mortality in acute pulmonary embolism (PE). Although clot burden has not been consistently shown to correlate with disease outcomes, proximally located PE is generally perceived as more severe. Purpose: To explore the ability of the Mean Bilateral Proximal Extension of the Clot (MBPEC) score to predict mortality and adverse outcome. Methods: This was a single center retrospective cohort study. 1743 patients with computed tomography pulmonary arteriography (CTPA) verified PE diagnosed between 2005 and 2020 were included. Patients with active malignancy were excluded. The PE clot burden was assessed with MBPEC score: The most proximal extension of PE was scored in each lung from 1 = sub-segmental to 4 = central. The MBPEC score is the score from each lung divided by two and rounded up to nearest integer. Results: We found inconsistent associations between higher and lower MBPEC scores versus mortality. The all-cause 30-day mortality of 3.9% (95% CI: 3.0-4.9). The PE-related mortality was 2.4% (95% CI: 1.7-3.3). Patients with MBPEC score 1 had higher all-cause mortality compared to patients with MBPEC score 4: Crude Hazard Ratio (cHR) was 2.02 (95% CI: 1.09-3.72). PE-related mortality was lower in patients with MBPEC score 3 compared to score 4: cHR 0.22 (95% CI: 0.05-0.93). Patients with MBPEC score 4 did more often receive systemic thrombolysis compared to patients with MBPEC score 1-3: 3.2% vs. 0.6% (p < .001). Patients with MBPEC score 4 where more often admitted to the intensive care unit: 13% vs. 4.7% (p < .001). Conclusion: We found no consistent association between the MBPEC score and mortality. Our results therefore indicate that peripheral PE does not necessarily entail a lower morality risk than proximal PE.

Recurrent pericarditis. / Residiverende perikarditt.

Pericarditis is an important differential diagnosis in patients with chest pain. The two most common causes in the developed world are idiopathic pericarditis and inflammation following cardiac surgery or myocardial infarction. Recurrence of pericarditis affects up to 30 % of patients, half of whom experience multiple episodes, and approximately 10 % develop steroid-dependent and colchicine-refractory pericarditis. Recurrence is due to autoinflammatory processes in the pericardium. Advanced diagnostic imaging and treatment with colchicine and interleukin-1 inhibitors has helped reduce morbidity considerably in recent years. In this clinical review, we summarise up-to-date knowledge about the diagnostic evaluation and treatment of patients with recurrent primary pericarditis.

Accuracy of stress perfusion cardiac magnetic resonance imaging in a district hospital.

Background: The European Society of Cardiology has published updated guidelines regarding pathways for diagnosis and management of obstructive coronary artery disease (CAD). Non-invasive functional assessment, for example, by stress perfusion cardiac magnetic resonance (stress pCMR) is recommended in patients with intermediate pretest probability of disease. Previous pCMR studies were mainly performed in high volume university hospitals with experienced radiologists or cardiologists interpreting the images. Purpose: The aim of the present study was to evaluate the feasibility of establishing a stress pCMR imaging service in a district hospital. Material and Methods: One hundred and thirteen patients with intermediate pretest probability of CAD referred for single-photon emission computed tomography (SPECT) at the regional hospital also underwent adenosine stress pCMR locally. The diagnostic analysis was compared to that of an experienced cardiac magnetic resonance (CMR) center serving as a reference. Results: Inter-rater agreement between local readers and the reference reader was substantial to perfect for late gadolinium enhancement (LGE) (weighted kappa = 0.76 and 0.82), but only fair to moderate for pCMR (k = 0.34 and 0.51). No improvement in agreement between reference reader and local reader during the study was demonstrated. Conclusion: CMR is feasible in patients with intermediate pretest probability of obstructive CAD in the setting of a district hospital. However, as opposed to infarct detection with LGE, the interpretation of stress pCMR was more challenging. To establish this method, we suggest obtaining experience in close collaboration with a reference CMR center.

The venous thrombosis registry in Østfold Hospital (TROLL registry) - design and cohort description.

Purpose: The incidence of venous thromboembolism (VTE) is expected to increase over the next decades, further increasing its substantial impact on patients and health care resources. Registries have the benefit of reporting real-world data without excluding clinically important subgroups. Our aim was to describe a Norwegian VTE registry and to provide descriptive data on the population and management. Registry Population: The Venous Thrombosis Registry in Østfold Hospital (TROLL) is an ongoing registry of consecutive patients diagnosed with, treated, and/or followed up for VTE at Østfold Hospital, Norway, since 2005. Baseline and follow-up data, including demographics, clinical features, risk factors, diagnostic procedures, classification of VTE, and treatment were collected during hospitalization, and at scheduled outpatient visits. Findings to Date: From January 2005 to June 2021, 5037 patients were eligible for research in TROLL. Median age was 67 years (interquartile range, 55-77), and 2622 (52.1%) were male. Of these, 2736 (54.3%) had pulmonary embolism (PE), 2034 (40.4%) had deep vein thrombosis (DVT), and 265 (5.3%) had upper-extremity DVT or splanchnic or cerebral sinus vein thrombosis. In total, 2330 (46.3%) were classified as unprovoked VTE, and 1131 (22.5%) had cancer. Direct oral anticoagulants were the most frequent therapeutic agents (39.3%) followed by low-molecular-weight heparins (30.4%) and vitamin K antagonists (30.3%). Outpatient treatment for PE increased from 4% in 2005 to 23% in 2019. Future Plans: TROLL is a population-based ongoing registry that represents a valuable source of real-world data that will be used for future research on the management and outcomes of VTE.

Prevalence of Sclerosing Cholangitis Detected by Magnetic Resonance Cholangiography in Patients With Long-term Inflammatory Bowel Disease.

BACKGROUND & AIMS: The prevalence of primary sclerosing cholangitis (PSC) among patients with inflammatory bowel disease (IBD) is unclear. Patients with IBD might be screened for PSC using magnetic resonance cholangiography (MRC). We aimed to estimate the frequency and distribution of MRC-detected lesions that indicate PSC in patients with IBD 20 years after their initial diagnosis and to identify clinical characteristics associated with these findings. METHODS: We performed a follow-up analysis of a population-based cohort of 756 patients in South-Eastern Norway diagnosed with IBD from January 1, 1990 through December 31, 1993. Of these subjects, 470 attended a follow-up evaluation 20 years later in which they were offered routine clinical blood testing and ileocolonoscopy; 322 were screened by MRC (222 with ulcerative colitis and 100 with Crohn's disease). Two radiologists independently evaluated results from the MRC examinations. RESULTS: In the MRC examination, 24 patients (7.5%) were found to have PSC-like lesions; only 7 of these patients (2.2%) were known to have PSC. One patient was initially missed and 1 had small-duct PSC, so the final prevalence of PSC was 8.1%. Extensive colitis, a high prevalence of colectomy, and chronic and continuous symptoms of IBD occurred in significantly more patients with suspected PSC than without PSC (P = .029, P = .002, and P = .012, respectively). Among patients with subclinical features of PSC, the MRC progression score for PSC increased when they were re-examined after a median 3.2 years (P = .046). CONCLUSIONS: Using MRC analysis of patients with long-term IBD, we found the prevalence of PSC to be around 3-fold higher than that detected based on symptoms. Sixty-five percent of patients had subclinical PSC associated with progressive IBD, with no biochemical abnormalities and mild disease, based on radiology findings. PSC appears to progress in patients with subclinical disease, but long-term outcomes are not known.