A recent study identified numerous non-traditional risk factors for cardiovascular disease in patients with type 2 diabetes mellitus, suggesting that current prediction models may overlook critical indicators.
In the study, published in Communications Medicine, investigators analyzed data from over 459,000 participants in the UK Biobank, stratifying them by their history of diabetes and cardiovascular disease (CVD).
The investigators revealed that classic CVD risk factors, such as family history and high blood pressure, were less significant in predicting CVD among patients with type 2 diabetes mellitus (T2D). Instead, the top predictors for those with T2D but no history of CVD included cystatin C, self-reported health satisfaction, and biochemical measures of poor health. Among patients with both T2D and a history of CVD, key predictors were self-reported poor health and blood cell counts.
"Non-traditional risk factors are of particular importance in [patients] with diabetes," said lead study author Katarzyna Dziopa, of the Institute of Health Informatics at the University College London, and colleagues. Adding these features could improve risk stratification; per 1,000 patients with diabetes, 133 CVD and 165 heart failure cases could receive a higher risk ranking.
The investigators analyzed over 600 features, ultimately identifying 382 relevant variables after data engineering steps. They employed a penalized generalized linear model with a binomial distribution to identify CVD-related features, followed by a 20% hold-out set to replicate identified features and provide an importance-based ranking.
Among the findings, unique diabetes-related risk factors included dietary patterns, mental health indicators, and specific biochemical measures such as estradiol and rheumatoid factors. The investigators noted that these factors could enhance the predictive accuracy of existing CVD risk models, which currently don't account for these variables.
The investigators also highlighted that the performance of CVD prediction models in patients with T2D was notably lower compared with the general population, with c-statistics of approximately 0.70 compared with 0.88 in women and 0.86 in men. This discrepancy underscored the necessity for tailored risk prediction models that incorporate non-traditional risk factors.
The study's findings are particularly relevant given the rising prevalence of T2D and its association with increased cardiovascular morbidity and mortality. Incorporating these non-traditional risk factors into clinical practice could lead to improved identification and management of high-risk patients with diabetes.
The authors reported no conflicts of interest.
