Risk for cardiovascular events responds nonlinearly to carotid intima-media thickness in the KORA F4 study

Background and aims: Risk assessment studies on the impact of carotid intima-media thickness (CIMT) on cardiovascular events (CVEs) often apply a linear relationship in Cox models of proportional hazards. However, CVEs are mostly induced through rupture of plaques driven by nonlinear mechanical properties of the arterial wall. Hence, the risk response might be nonlinear as well and should be detectable in CVE incidence data when associated with CIMT as surrogate variable for atherosclerotic wall degeneration. Methods: To test this hypothesis, we investigate the KORA F4 study comprising 2580 participants with CIMT measurements and 153 first CVEs (86 strokes and 67 myocardial infarctions). CIMT is only a moderate predictor of CVE risk due to confounding by attained age. Biological evidence suggests that age-related CIMT growth is not entirely connected with atherosclerosis. To explore the complex relations between age, CIMT and CVE risk, we apply linear and nonlinear models of both CIMT and dnCIMT, defined as deviation from a sex and age-adjusted normal value. Results: Based on goodness-of-fit and biological plausibility, threshold and logistic step models clearly reveal nonlinear risk response relations for vascular covariables CIMT and dnCIMT. The effect is more pronounced for models involving dnCIMT as novel risk factor, which is not correlated with age. Conclusions: Compared to the standard approach of risk assessment with linear models involving CIMT, the application of excess dnCIMT with nonlinear risk responses leads to a more precise identification of asymptomatic high risk patients, especially at younger age.