In a large community-based sample, we provide absolute CVD risk estimates for the entire sample and stratified by age and by sex for different categories of LDL levels, representing different treatment intensities of LLT. Furthermore, we assessed potential mechanisms by which the residual risk could be mediated: subclinical carotid disease, neurohormonal activation or chronic inflammation. In addition, we describe the subclinical disease burden at the carotid arteries associated with different lipid levels in the community, stratified by treatment status.
Principal observations
First, subclinical disease burden and risk for incident CVD rose with increasing LDL category. Second, individuals who reach LDL levels <130 mg/dL on treatment (group 3) as well as individuals with LDL ≥130 mg/dL (groups 4 and 5) displayed greater carotid atherosclerotic disease burden, as compared with untreated individuals with usual LDL <100 mg/dL (group 1). Third, individuals on lipid-lowering medication (groups 5 and 3) had a substantial residual absolute risk for incident CVD that was about three times the risk observed in individuals with LDL <100 mg/dL who were not on treatment. Residual CVD risk rose substantially with age, and men had slightly higher absolute CVD risk than women in most LDL categories. Fourth, part of this increased risk was due to worse cardiovascular risk profiles in groups 3–5 (table 1). However, on adjustment for traditional CVD risk factors, individuals on LLT who reach LDL <130 mg/dL still had about 50% greater hazards for CVD as compared with the referent group. Similarly, groups 4 and 5 displayed statistically significant HRs for CVD (compared with group 1), taking potential confounders into account. Fifth, adjustment for subclinical carotid disease burden only moderately attenuated these HRs (by about 5%) consistent with partial mediation of the residual CVD risk by the burden of subclinical atherosclerosis. Further adjustment for CRP and BNP did not significantly alter the risk estimates.
In the context of the published literature
The association between lipid measures and CVD risk is well established in the epidemiological and clinical literature.17 Numerous clinical trials have demonstrated that lowering of LDL-C through pharmacological or lifestyle interventions reduces the risk for recurrent CVD and mortality in patients with established CVD1 18 or in the primary prevention setting.2 19 20 Furthermore, clinical trials have shown that statins—by lowering LDL—also improve subclinical CVD measures. A recent meta-analysis of 11 clinical trials reported beneficial effects of statins on carotid atherosclerosis.21
Despite clinically and statistically significant reductions in clinical and subclinical CVD risk in the LLT arm compared with the respective comparison groups, many clinical trials reported a relevant residual cardiovascular risk in individuals receiving LLT, with the magnitude of the residual risk depending on the exact patient sample, the duration of follow-up, the type (generation) and dosage of treatment used, the target LDL levels that have been reached, and the exact endpoint investigated in the study.1 18 19 22–25 Clinical trials have relatively strict inclusion and exclusion criteria so that it is unclear whether the observations from clinical trials regarding the residual cardiovascular risk on LLT are applicable to the general population.26
In our unselected sample from the community, we confirm a substantial residual absolute CVD risk in individuals on LLT (group 3 and group 5). Individuals in these groups had about three times the risk of the referent group 1 (untreated LDL <100 mg/dL) in the overall sample. In individuals below 60 years of age, the ratios of the absolute disease risks in those on LLT and those in the referent group 1 were even more pronounced.
Furthermore, we explored the subclinical disease burden at the carotid arteries associated with various LDL levels, stratified by LLT status. Individuals on LLT and with greater LDL levels displayed statistically significantly greater odds for carotid atherosclerosis on ultrasound, including increased IMT and stenoses,13 consistent with other observational studies.27 28 In addition, we evaluated to what extent the increased hazards for CVD in LDL groups 3–5 (figure 3) were attributable to subclinical atherosclerosis at the carotid arteries and to systemic inflammation and neurohormonal activation. Our observations indicate that adjustment for subclinical carotid atherosclerosis only modestly attenuated the relative risk estimates, consistent with the concept that a modest proportion of the increased risk is attributable to subclinical atherosclerosis. Additional adjustment for established biomarkers of systemic inflammation (CRP) and neurohormonal activation (BNP) had no significant effect on the risk estimates.
Important drivers for the elevated subclinical and clinical CVD risk in individuals on LLT with LDL levels <130 mg/dL (group 3) are unfavourable levels for other risk factors, including systolic blood pressure, body mass index, diabetes and HDL in participants in this group (table 1). The greater CVD risk at baseline in individuals on LLT (also reflected in online supplementary figure 1) may reflect indication bias (confounding by indication; higher risk individuals are more likely to be treated by their physicians), as expected in observational (non-randomised) studies assessing residual risk on treatments.29 Indeed, one of the factors contributing to the decision whether LLT is initiated in a given patient is the overall CVD risk over a 10-year time horizon, which takes levels of several risk factors into account.17 On the other hand, and as a clinical consequence, these observations underscore the importance of monitoring all traditional risk factors in individuals on LLT, even if people reach LDL levels below 130 mg/dL.
It also has to be kept in mind that LDL-cholesterol, despite being an established lipid measure and CVD risk factors, might mirror LDL-related residual CVD risk not ideally, particularly in subgroups of the population, where other lipid measures (such as apolipoprotein B (ApoB) concentrations) provide discordant information. In the Women’s Health Study, approximately one in five women were discordant for LDL-C and ApoB; and CVD risk in these individuals was underestimated or overestimated, when estimations were solely based on LDL-C.30 Also in the Framingham Offspring cohort, ApoB levels provided additional information about CVD risk, beyond established lipid measures, in individuals, where the observed ApoB levels were discordant to the expected ApoB concentrations, as predicted based on circulating LDL-C.31 Furthermore, in Jupiter trial participants (individuals with LDL <130 mg/dL and high-sensitivity C-reactive protein (hsCRP) ≥2.0 mg/L), ApoB levels, but not LDL-C, were associated with incident CVD in the placebo arm.32 In the on-statin arm, however, ApoB failed to reach significance (HR: 1.20, 95% CI 0.98 to 1.46 for incident CVD).32
Strengths and limitations
The strengths of our investigation include the large community-based sample, the careful and comprehensive characterisation of our study participants, including measures of subclinical atherosclerosis at the carotid arteries and of biomarkers of systemic inflammation and neurohormonal activation, and the prospective follow-up for incident CVD (median follow-up, 13.7 years). The following limitations merit consideration. We assigned participants to LDL groups based on a single LDL measurement and a single assessment of the intake of lipid-lowering medication. This might result in some misclassification, which is likely to be non-differential; this would bias us towards the null hypothesis of no difference between the groups in risk of outcome events. In our analyses we did not differentiate between different types of lipid-lowering medication, dosages and durations of treatment. Unfortunately, information on the generation of statin taken or on statin dose was not available in our sample. Furthermore, the prescription of LLT is determined by circulating lipid concentrations and by the overall burden of CVD risk factors. Thus, individuals on LLT in the community might a priori be at higher CVD risk as compared with individuals not receiving such medication (confounding by indication).29 Finally, we have a relatively small number of individuals on statins. However, these individuals were well phenotyped and had a relatively long follow-up period well beyond that of typical clinical trials of LLT.
In conclusion, we observed a substantial residual cardiovascular risk in individuals on lipid-lowering medication in the community, partly explained by an adverse profile of other CVD risk factors and in part (by about 5%) by carotid subclinical atherosclerosis in these people. These observations emphasise the necessity to closely monitor all standard CVD risk factors in individuals on LLT, even when LDL levels are lowered effectively.