Compared to the middle-aged, in the elderly, the evidence of an association between lipids and subsequent coronary heart disease (CHD) in observational studies is less clear [1-4], although statins have been shown to reduce the risk of cardiovascular disease in this population [5,6]. Since the late 1990s, statin therapy has been increasingly prescribed for the elderly population [7,8]. The possible link between lipids and CHD risk in the elderly thus needs to be revisited while separating those receiving lipid-lowering treatment (LLT) from those who do not, a study that has not yet been undertaken. In this light, using data from the Three-City Study, a large French population based prospective study on community-dwelling elderly [9] we attempted to clarify the association between lipids, apolipoproteins, and CHD risk in the elderly, by separating individuals based on the prescription of LLT and statin therapy in particular, in the primary prevention setting.
Methods
The Three-City Study is a French multicenter prospective cohort investigating the determinants of CHD, stroke, and dementia in community dwelling older participants [9]. Briefly, 9,294 men and women aged ≥ 65 years were selected from the electoral rolls of 3 large cities and agreed to participate (3,649 men and 5,825 women). The study protocol was approved by the Ethics Committee of the University Hospital of Kremlin-Bicêtre, and each participant signed an informed consent.
A face-to-face questionnaire on health was administered to participants who underwent a physical examination too. The questionnaire included an inventory of all drugs used during the preceding month. No data on doses and duration were collected. The concordance rate between self-reported use of LLT and individual data extracted from the database of the French National Health Insurance System was 92% for a subsample of the 3C Study [10].
Blood was collected at baseline after overnight fasting. Total cholesterol, HDL-cholesterol, triglycerides (TG), and glycemia were measured in fresh samples in the entire cohort in a single laboratory using standard methods. LDL-cholesterol was estimated by the Friedewald’s formula for TG ≤ 4.5 g/L. Non-HDL cholesterol was calculated as the difference between total cholesterol and HDL cholesterol. Diabetes was defined by a fasting glycemia greater than 126 mg/dl or by the use of antidiabetic medications.
A case cohort study was built for the investigation of non-standard risk markers after 4 years of follow-up. It comprised 198 subjects who experienced a first CHD event during the 4 years of follow up and a sub-cohort of 1,081 subjects randomly selected from the initial cohort in strata defined by center, age (5-year intervals) and gender. These subjects had all lipid measures available and had no personal history of CHD at baseline. Apolipoprotein A-I (ApoA), apolipoprotein B-100 (ApoB) and high-sensitivity CRP (hs-CRP) were centrally measured in thawed baseline plasma by technicians blind to CHD status using a particle-enhanced turbidimetric immunoassay. CHD events were surveyed during intermediate examinations at the study center two and four years after baseline examination, were the subjects were interviewed about the occurrence of any hospitalizations for a CHD event including stable and unstable angina pectoris, coronary balloon dilatation, or artery bypass, myocardial infarction, and death due to CHD. For all event reported, a detailed enquiry was made by trained physicians through hospital charts, interview with the patient's physician or the family, and in case of a fatal event, death certificates, and autopsy reports. Events were finally adjudicated by an independent event validation committee.
In participants who did not receive LLT, those who received statins, and those who received fibrates, a Cox proportional hazard model suited to the case-cohort design was used to estimate the standardized hazards ratio (HRs) and 95% confidence intervals (95% CIs) of each lipid parameter with respect to CHD [11]. Analyses were first adjusted for age, gender, study center, and educational level (model 1) and then for daily total alcohol use, body mass index (BMI), diabetes, systolic blood pressure, antihypertensive treatment and hs-CRP (model 2).
Results
The sub-cohort consisted of 414 men (38.1%), with a mean age (± SD) of 73.7 years (± 5.3). There were 816 participants (75.1%) who did not receive any LLT, 145 (13.4%) who received fibrates, and 125 (11.5%) who received statin therapy. Compared to the non-treated group and with respect to non-lipid risk factors, those under statins tended to be more diabetic (p = 0.07) and to have a higher BMI (p = 0.08), whereas those under fibrates were more likely to be women (p = 0.07), were more diabetic (p = 0.02) and tended to have higher hs-CRP levels (p = 0.09).
Among participants who were non-users of LLT at baseline, 145 had a first CHD event during the follow-up period. In this group, as shown in Table 1, LDL cholesterol, triglycerides, non-HDL cholesterol, ApoB100, ApoB/ApoA1, and LDL/HDL cholesterol ratios were all predictive of CHD after adjustment for age, gender, study center, and educational level; ApoA1 and HDL cholesterol had a marginally significant inverse association with CHD. After additional adjustment (model 2) these lipids but TG, ApoA1, and HDL cholesterol, remained predictors of CHD. In addition, these associations were of comparable strength.
Among individuals receiving statins at baseline, 27 had a CHD event during follow up. In this group, most lipid parameters, except ApoA1 and HDL cholesterol, were significant predictors of CHD after adjustment for age, gender, study center, and educational level. The hazard ratios increased after adjustment for the confounders (model 2) except for triglycerides and ApoA1. Moreover, for a given lipid parameter, the association with CHD was stronger in those under statin therapy than in those who were non-users of LLT.
Among individuals on fibrates at baseline, 27 developed CHD during follow up. In this group, triglycerides and HDL-cholesterol were significant predictors of CHD after adjustment for age, gender, study center, and educational level; after fully adjusted analyses, only triglycerides remained associated with CHD.
Discussion
In contrast to the middle-aged population, there remains some uncertainty regarding the association between lipids and incident CHD in the elderly. In survivors of the Whitehall Study comprising elderly men aged 77 years on average, higher than usual levels of total cholesterol, LDL cholesterol and ApoB were associated with higher mortality due to ischemic heart disease over 7 years in those without previous cardiovascular disease (75% of the study cohort) (3). In the Cardiovascular Health Study, which included men and women aged 65 years and over, total and LDL cholesterol and triglycerides, were weakly and often marginally associated with cardiovascular disease, (myocardial infarction, stroke and overall mortality) during the 7.5 years of follow up [4].
The current analysis extends the results of these previous studies by separately investigating those who received LLT at baseline and those who did not. All lipids except ApoA1 and HDL cholesterol were independent predictors of CHD in those not taking any LLT at baseline, indirectly corroborating the benefits of statin therapy in reducing cardiovascular disease in elderly participants of the PROSPER and the AFCAPS/TexCAPS trials [5,6]. We also found that lipids remained independent predictors of CHD under statin therapy, possibly with a stronger association than that observed in the absence of LLT. This result may reflect a lack of adherence to statin treatment, which in itself may increase the risk of CHD [12]. A previous analysis of the Three-City Study indicated, however, that drug exposure and compliance with treatment were fairly good [10]. Alternatively, subjects under statin therapy may have a worse cardiovascular risk profile or an underlying chronic condition (indication bias). However, the results remained unchanged when adjustment for baseline characteristics was carried out. Therefore, the current study suggests that under statin therapy, there remains a substantial residual risk of CHD that is attributable to some extent to lipid levels.
Moreover, the current study indicates that the pattern of association between lipids and CHD is different between those under statins and those under fibrates. This heterogeneity might reflect differences in the baseline characteristics between statin and fibrate users. However, adjustment for these differences did not eliminate such heterogeneity. It may also reflect the indication for which the respective agent was given, usually hypercholesterolemia for statins and hypertriglyceridemia for fibrates.
The current results should be interpreted keeping in mind the limited numbers of CHD events, especially in those receiving LLT. Moreover, information on the duration and titration of the treatment were unavailable at baseline examination. Finally, the allocation of LLT was not controlled randomly and may have been subject to an indication bias.
To conclude, in community-dwelling elderly participants from the Three-City Study, there was some heterogeneity in the association between lipids and future CHD according to the lipid-lowering treatment received at baseline. In those who did not receive any lipid-lowering treatment, LDL cholesterol, non-HDL cholesterol, ApoB100, ApoB/ApoA1, and LDL/HDL cholesterol ratios were all predictors of CHD over 4 years, independently of other traditional risk factors and hs-CRP. Among those who received statin therapy (11.5% of the study cohort), the same lipid parameters were predictors of CHD, but in general with greater strength. In those who were on fibrates (13% of the study cohort), only triglycerides were independent predictors of CHD.