Elevated remnant cholesterol and atherosclerotic cardiovascular disease in diabetes: a population-based prospective cohort study

We included 107,243 white Danish individuals aged 20–100 years with information on lipid levels and statin use from the Copenhagen General Population Study, which consists of individuals randomly recruited from the Copenhagen general population in 2003–2015. Among these individuals, 4569 had diabetes as defined by the criteria below. At baseline, participants completed a questionnaire, underwent a physical examination and had blood drawn for biochemical analyses. The study was approved by a local institutional review board and a Danish ethical committee (H-KF-01-144/01). All participants signed written informed consent, and the study was conducted in accordance with the Declaration of Helsinki.

Individuals were defined as having diabetes if there was a diagnosis of diabetes mellitus type 1 or 2 (ICD-8 codes 249–250 and ICD-10 codes E10–E14 [http://​apps.​who.​int/​classifications/​icd10/​browse/​2016/​en]) in the national Danish Patient Registry before the baseline examination, or on the basis of self-reported diabetes, use of glucose-lowering medications or a non-fasting plasma glucose >11 mmol/l (198 mg/dl) at the baseline examination.

Individuals living in Denmark have since 1968 been assigned a Central Person Registration number at birth or immigration through the national Danish Civil Registration System, allowing linkage to nationwide Danish health registries, ensuring no losses to follow-up.

Non-fasting plasma triglycerides, total cholesterol, HDL-cholesterol, high-sensitivity C-reactive protein and glucose were measured using standard hospital assays. LDL-cholesterol was calculated using the Friedewald equation [11], unless triglycerides were >352 mg/dl (4 mmol/l), where it was measured directly using a standard hospital assay. Remnant cholesterol was total cholesterol minus LDL-cholesterol minus HDL-cholesterol. In 1906 of the individuals with diabetes, remnant cholesterol, LDL-cholesterol and LDL triglycerides were measured directly using NMR spectroscopy [12]. Remnant cholesterol was the cholesterol in all VLDL subfractions using this method.

Information on ethnicity, gender, smoking and medication use was self-reported at baseline and were not tracked during follow-up. Use of lipid-lowering medications was >99% statins. Smoking status was recorded as never, former or current smoking, and cumulative smoking as number of pack-years smoked. Date of birth and sex were obtained from the Danish Civil Registration System. Weight and height for calculation of BMI and BP were measured at the baseline examination. Glucose-lowering medication use was insulin, other glucose-lowering medication, or both insulin and other glucose-lowering medication.

More detailed description of the statistical analysis is given in the electronic supplementary material (ESM) Methods.

We followed 4569 individuals with diabetes from the Copenhagen General Population Study for up to 15 years (median 8 years), during which time 236 individuals were diagnosed with peripheral artery disease, 234 with myocardial infarction, 226 with ischaemic stroke and 498 with any ASCVD (some individuals have multiple events). Baseline characteristics by remnant cholesterol levels are shown in Table 1, by statin use in ESM Table 1, by smoking status in ESM Table 2 and by diabetes type in ESM Table 3.

Compared to 102,674 individuals without diabetes, individuals with diabetes had higher remnant cholesterol levels but lower LDL-cholesterol levels (Fig. 1). In individuals with diabetes, remnant cholesterol levels were similar regardless of statin use, while LDL-cholesterol levels were lower in individuals using statins compared with those not using statins (Fig. 1). Remnant cholesterol and LDL-cholesterol were weakly correlated (R²=2.4%; ESM Fig. 1).

In individuals with diabetes, multivariable adjusted HRs (95% CI) for risk of peripheral artery disease were 1.6 (1.1, 2.3) per doubling of remnant cholesterol and 0.9 (0.6, 1.2) per doubling of LDL-cholesterol (Fig. 2). Corresponding HRs were 1.8 (1.2, 2.5) and 1.0 (0.7, 1.4) for myocardial infarction, 1.5 (1.0, 2.1) and 1.1 (0.8, 1.6) for ischaemic stroke, and 1.6 (1.2, 2.0) and 0.9 (0.7, 1.1) for any ASCVD, respectively. These results were largely similar in statin users and non-users alike and in non-smokers and smokers alike (ESM Fig. 2 and ESM Fig. 3).

Results were similar after additional adjustment for BMI and ASCVD before baseline, in competing risk regression with death as competing event, when using time in study as timescale, when additionally adjusting for high-sensitivity C-reactive protein, glucose-lowering medication use, BMI and HDL-cholesterol, when using different equations to calculate LDL and remnant cholesterol, and after excluding individuals with type 1 diabetes (ESM Fig. 4–8). As expected, estimates were attenuated without correction for regression dilution bias (ESM Fig. 4). When additionally adjusting for apolipoprotein B, estimates were similar for risk of ischaemic stroke and any ASCVD, attenuated for risk of peripheral artery disease, and strengthened for risk of myocardial infarction (ESM Fig. 6). When adjusting for plasma triglycerides, the confidence intervals were wider, as expected because of high correlation between remnant cholesterol and triglycerides (ESM Fig. 6). After excluding individuals with ASCVD before baseline, the association between elevated remnant cholesterol and risk of peripheral artery disease was slightly attenuated (ESM Fig. 8). When adjusting for LDL triglycerides measured in about half of the included individuals, estimates except for risk of myocardial infarction attenuated to the null (ESM Fig. 9). Estimates were attenuated in individuals with non-fasting plasma glucose below the median (ESM Fig. 10). Directly measured remnant cholesterol was less strongly associated with risk of peripheral artery disease and myocardial infarction but had a similar association with ischaemic stroke (ESM Fig. 11). Results for directly measured LDL-cholesterol were similar to calculated LDL-cholesterol (ESM Fig. 11). Results for directly measured remnant cholesterol were attenuated to the null after adjustment for LDL triglycerides; likewise, LDL triglycerides were not associated with risk of ASCVD after adjustment for directly measured remnant cholesterol (ESM Fig. 12). Results were similar in men and women (ESM Fig. 13).

In all individuals, the HR (95% CI) for risk of peripheral artery disease was 2.5 (2.2, 2.9) for diabetes vs no diabetes (Fig. 3). Corresponding HRs were 1.6 (1.4, 1.9) for myocardial infarction, 1.4 (1.2, 1.6) for ischaemic stroke, and 1.6 (1.5, 1.8) for any ASCVD. Results were similar in statin non-users, non-smokers (Fig. 3) and in smokers (ESM Fig. 14). In statin users, all HRs were attenuated (ESM Fig. 14).

In individuals with diabetes, levels of remnant cholesterol and high-sensitivity C-reactive protein were higher compared to individuals without diabetes (ESM Fig. 15). Levels of LDL-cholesterol were lower in individuals with diabetes, also when comparing only within statin non-users or non-smokers and within statin users or smokers (ESM Fig. 15). LDL-cholesterol levels remained lower also after additional adjustment for age, sex, smoking status and cumulative smoking (data not shown). Because of this, elevated LDL-cholesterol was not tested for in the analysis of explained excess risk.

Individuals with diabetes commonly have elevated remnant cholesterol levels, probably as a result of both increased production of triglyceride-rich lipoproteins due to dietary and lifestyle factors, and increased retention of triglyceride-rich lipoproteins in plasma due to insulin resistance and hyperglycaemia, which inhibit lipoprotein lipase activity [8, 15]. Remnant cholesterol in triglyceride-rich lipoproteins can penetrate the arterial intima and accumulate in atherosclerotic plaques alongside LDL [3, 16], and elevated remnant cholesterol is a causal risk factor for ASCVD [2, 3]. Indeed, most cases of peripheral artery disease and myocardial infarction result from atherosclerosis [17], as do many cases of ischaemic stroke [18]. Additionally, our results in individuals with high non-fasting plasma glucose indicate that elevated remnant cholesterol may explain increased ASCVD risk conferred by postprandial lipaemia or poor glycaemic control [19]. This may be because postprandial lipaemia better reflects cumulative exposure to high levels of triglyceride-rich lipoproteins compared with fasting lipaemia, as most of the day is spent in a postprandial state [20], or because postprandial lipaemia reflects adverse lifestyle factors including poor diet, alcohol consumption and smoking [21].

In support of some of our findings, studies of general population cohorts [4‐6, 32, 33] corroborate the association between elevated remnant cholesterol and increased risk of peripheral artery disease, myocardial infarction, ischaemic stroke and any ASCVD in individuals with diabetes. Additionally, some studies have found associations between elevated remnant cholesterol and increased risk of coronary artery disease and cardiovascular death in individuals with diabetes and prediabetes [34‐36]. Taken together, in the present study we demonstrate in a contemporary population with access to modern primary prevention that elevated remnant cholesterol explained 24% of the excess risk of ASCVD in those with diabetes vs those without diabetes, while LDL-cholesterol did not explain any excess risk of ASCVD.

The relatively large size of this diabetes cohort, long follow-up without losses, and large number of events collected from nationwide registers led to enough statistical power for clinically important stratified analyses. Other strengths include that Danish registry diagnoses of myocardial infarction have high diagnostic accuracy [37] and that ischaemic stroke diagnoses were validated in each patient individually [5]. Further, standard laboratory measurements were used to calculate remnant cholesterol, making the results easily applicable in a clinical setting. Additionally, our study period started at a time when fewer individuals with diabetes received statin treatment than now, allowing us to compare individuals with and without statin treatment at baseline. Finally, it is a strength that the relatively high use of statins during the study means that our results are likely to be generalisable to other contemporary populations in high-income countries.

Observational studies are always limited by potential confounding and reverse causation, which means causal conclusions are difficult to draw. Although we adjusted for important known risk factors, residual and unmeasured confounding cannot be completely excluded. One potential unobserved confounder is changes in statin and glucose-lowering treatment after the baseline examination, as described above. Another limitation is that we could not adjust for reliable markers of glycaemic control (fasting glucose or HbA_1c) since they were not measured, and therefore we only adjusted for non-fasting glucose, which may reflect postprandial lipaemia, poor glycaemic control or both.

We observed that individuals with diabetes still have high excess risk of ASCVD, especially peripheral artery disease, despite access to statins, glucose-lowering drugs and other medications. Importantly, our results do not contradict the need of achieving optimal control of LDL-cholesterol for prevention of ASCVD in individuals with diabetes. However, excess risk of ASCVD in populations with high use of statins could be caused by other factors, including elevated remnant cholesterol and low-grade inflammation.

A doubling of remnant cholesterol in individuals with diabetes was associated with a 1.6-fold increased risk of ASCVD, which included 1.6-fold risk of peripheral artery disease, 1.8-fold risk of myocardial infarction and 1.5-fold risk of ischaemic stroke. Elevated remnant cholesterol explained 24% and low-grade inflammation explained 18% of the excess risk of ASCVD in diabetes. Clinical trials should test if substantial remnant cholesterol-lowering therapy without increases in LDL-cholesterol and apolipoprotein B can prevent ASCVD in diabetes.