Individualized Cardiovascular Disease Prevention: Prediction of risk and treatment effect

Abstract

In the present era of evidence-based medicine, the study of groups is the dominant paradigm to establish causes of disease and to determine the efficacy of treatment. The results of these studies are usually presented as average group-level estimates, which are not informative of the individual patient’s effect. Clinicians are faced with the challenge to translate the evidence from such large (intervention) studies to the care for individual patients in clinical practice.

Blood pressure-lowering, cholesterol-lowering and antiplatelet therapy can reduce the risk of adverse cardiovascular outcomes at a group level. However, some patients will benefit more than average, while others do not benefit or may even be harmed. We used data from large randomized controlled trials to develop multifactorial prediction models to estimate the absolute effect of treatment for individual patients. In a large clinical trial evaluating the effect of routine blood pressure-lowering treatment and intensive glucose control in type 2 diabetes, we demonstrated a wide distribution in treatment effect for individual patients. A proportion of 43% of trial participants was identified to have a large 5-year absolute risk reduction of cardiovascular events with blood pressure (BP)-lowering treatment. On the other hand, the absolute effect of BP-lowering was small in 17% of patients. Intensive glucose control reduced the risk of vascular events but increased the risk of severe hypoglycaemia at a group level. We quantified the beneficial and adverse effects of treatment at a patient-specific level and showed that the majority of patients had a positive net effect of intensive glucose control. Further, we investigated the individual effects of angiotensin converting enzyme-inhibition in a large clinical trial in patients with stable coronary artery disease. A prediction algorithm based on clinical and genetic characteristics was able to identify 27% of patients with a zero or adverse treatment effect, whereas 28% of patients had a large absolute effect of treatment.

Further, we evaluated the comparative performance of ten cardiovascular risk scores for patients with type 2 diabetes in three different cohorts. Discriminative performance was generally modest but calibration was fairly accurate after adjusting the scores to the disease incidence of the target populations. These recalibrated models can assist clinicians to identify patients with diabetes who are at low or high cardiovascular risk. Next, we evaluated whether novel biomarkers could improve predictive utility. The addition of four novel biomarkers resulted in better discrimination and risk classification of patients with diabetes, although the improvements were modest.

Lastly, we evaluated heterogeneity in cardiovascular risk in patients with vascular disease and obesity. Obese patients who showed only few metabolic abnormalities were not at increased risk of recurrent cardiovascular disease. By contrast, lean patients with metabolic abnormalities were at high risk of recurrent vascular events. Differences in metabolic consequences of obesity might be partially explained by variations in body fat distribution. We observed a higher rate of visceral fat accrual in postmenopausal women compared with similar aged men, potentially contributing to an unfavourable metabolic profile and an increased risk of cardiovascular events in postmenopausal women.