Новые модели точнее прогнозируют риск сердечно-сосудистых заболеваний у пациентов с жировым гепатитом

New models based on accelerated failure time (AFT), compared with the widely used QRisk3 algorithm, more accurately predict risk for cardiovascular disease (CVD) among patients with metabolic dysfunction-associated steatohepatitis (MASH), according to study findings published in Diabetes, Obesity and Metabolism. Patients with MASH, estimated to affect 3% to 6% of the US population, have greater risks of CV morbidity and mortality, possibly related to systemic low-grade inflammation, oxidative stress, or other mechanisms. No CVD prediction models, however, have been specifically developed for the MASH population. In this retrospective cohort study, researchers examined information on 10,461 adult patients (mean age, 59.1 years; 51% women) with MASH and no history of CVD who were included in the UK Clinical Practice Research Datalink, which contains electronic medical records related to primary and secondary care. Over a mean follow-up of 5.04 years, 16.6% of patients had a CVD event. Several factors were predictors of CVD among both men and women, including age, body mass index (BMI), ethnicity, diabetes, rheumatoid arthritis, atrial fibrillation, chronic kidney disease, migraine, corticosteroid use, serious mental illness, cholesterol ratio, and variation in systolic blood pressure (BP). Additional sex-specific predictors were systolic BP level and erectile dysfunction among men and smoking and deprivation level among women. Using AFT models, the researchers developed new risk-prediction tools dubbed MASH-QRisk3-A, B, C, and D, and compared them with the original QRisk3 algorithm, which was designed for use in the general population. When examining calibration plots for predicted vs observed survival (in this case, a CVD event) at 1, 5, and 10 years, they found a “reasonable fit” for the new models. The researchers also identified a similar predictive value for time to CVD across the MASH-QRisk3 models, with C-statistics of about 0.70 in women and 0.72 in men, indicating moderate discrimination. Those figures indicated a predictive value in the MASH population lower than observed with specialized subsets, such as patients with type 1 diabetes, where the QRisk3 tool demonstrates an unadjusted C-statistic of 0.83 to 0.853, but were completely comparable to other real-world type 2 diabetes subpopulation scores. A comparison of the MASH-QRisk3 models and the QRisk3 algorithm in this MASH cohort, however, revealed greater accuracy with the AFT-based models for predicting CVD risk. “This modeling revealed that QRisk3 would both under- and overestimate the risk of a cardiovascular event in MASH patients with a maximum error of 12% in males and 13% in females, whereas the MASH-QRisk3 models (A–D) would underestimate the risk with a maximum error of 9% in males and 5% in females,” the researchers reported. Study limitations include uncertain generalizability of the new models outside the UK, possible unobserved confounding and missing information, reliance on clinical coding in the database, possible misclassification of MASH, lack of routine histological confirmation of MASH, use of a smaller derivation cohort for the MASH-QRisk3 models than for the QRisk3 algorithm, and a lack of comparator prediction tools other than QRisk3. “We describe a first-in-kind predictive model to assess the risk of CVD in patients with MASH, which has the potential to more accurately inform the treatment and management of this population,” the study authors concluded. Disclosure: This study was funded by Novo Nordisk. Multiple study authors declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of authors’ disclosures. References: Hollinghurst J, Augusto M, Tefos F, et al. Cardiovascular disease risk prediction in patients with metabolic dysfunction-associated steatohepatitis. Diabetes Obes Metab. Published online May 31, 2026. doi:10.1111/dom.70687