The Association between LDL-C and Incident Atherosclerotic Cardiovascular Disease in Older Adults: Results from the National Institutes of Health Pooled Cohorts

Abstract

BACKGROUND/OBJECTIVES:

Elevated low-density lipoprotein cholesterol (LDL-C) in early adulthood is associated with increased risk of atherosclerotic cardiovascular disease (ASCVD). The strength of the association between LDL-C and ASCVD among older adults, however, is less understood.

DESIGN:

We examined individual-level cohort data from the National Institutes of Health Pooled Cohorts (Framingham, Framingham offspring, MESA, & CHS), which prospectively measured CVD risk factors and incident disease.

SETTING:

Prospective cohort study.

PARTICIPANTS:

Adults ≥75 years free of ASCVD.

MEASUREMENTS:

We evaluated the associations between LDL-C and incident ASCVD (stroke, myocardial infarction, cardiovascular death) in unadjusted analysis and in multivariable adjusted Cox proportional hazards models. We assessed 5-year Kaplan-Meier ASCVD event rates in patients with and without hyperlipidemia (LDL-C ≥130 mg/dL or on lipid-lowering medications) stratified by the number of other risk factors including smoking, diabetes, and hypertension.

RESULTS:

We included 2667 adults ≥75 years (59% female) free of ASCVD; median age was 78 years with median LDL-C 117 mg/dL. In both unadjusted and adjusted analyses, there was no association between LDL-C and ASCVD (adjusted hazard ratio 1.022, 95% confidence interval 0.998–1.046, p=0.07). Among adults without other risk factors (free of smoking, diabetes, and hypertension), event rates were similar between those with and without hyperlipidemia (Kaplan-Meier rate 5.8% and 7.0%, respectively). Among adults with 1 or 2+ other risk factors, the presence of hyperlipidemia was also not associated with 5-year CVD event rates (Kaplan-Meier rate 12.8% vs. 15.0%, p=0.44 for 1 other risk factor, 21.9% vs. 24.0%, p=0.59 for 2+).

CONCLUSION:

Among a well-characterized cohort, LDL-C was not associated with CVD risk among adults aged 75 years or older, even in the presence of other risk factors.

INTRODUCTION

Hyperlipidemia is a well-established modifiable risk factor for cardiovascular events among adults.^1–3 In population-wide studies, for every 1 mmol/L increase in non–high-density lipoprotein cholesterol, ischemic heart disease mortality increases by approximately one-third⁴; however, much of this evidence comes from research on younger populations. The association between other lipoprotein laboratory parameters such as total cholesterol and mortality has been previously evaluated in older adults^4–6. A few epidemiological studies have assessed the association between low-density cholesterol (LDL-C) and mortality, but those including older adults have found conflicting results^4,6–11. Even fewer studies have examined the association of low-density lipoprotein cholesterol (LDL-C) and atherosclerotic cardiovascular disease (ASCVD) risk, including cardiovascular events in addition to cardiovascular mortality, in those with advanced age; those that have were limited by small sample sizes taken from a single city. ^6,8 As a result, whether LDL-C levels can be used to risk-stratify older adults who survive to older ages without actually developing ASCVD is uncertain.

In addition to few epidemiological studies, there are also few randomized primary prevention trials of statin therapy for ASCVD that have included older adults (≥75 years old).^12–15 The PROSPER (Prospective Study of Pravastatin in the Elderly at Risk) trial included 5804 patients, 70–82 years old, with known vascular disease or with risk factors for vascular disease, and found that pravastatin therapy reduced the risk of cardiovascular events compared with placebo.¹⁴ However, a sub-group analysis of the primary prevention cohort (N=3239) found no significant improvement in the primary endpoint of CHD death, non-fatal MI and fatal or non-fatal stroke.¹⁴ A sub-group analysis of the JUPITER (Justification for Use of Statins in Prevention: An Intervention Trial Evaluating Rosuvastatin) trial included 5595 patients 70 years or older and did find a 39% reduction in the hazard of the primary end point of MI, unstable angina, stroke, arterial revascularization or cardiovascular death with rosuvastatin vs. placebo.¹⁵ More recently, the results of a randomized trial investigating the efficacy of ezetimibe in high-risk primary prevention Japanese older adults (≥75 years old) were presented as a late breaking clinical trial at the American Heart Association 2018 Scientific Sessions, demonstrating a significant reduction in cardiovascular events at 5-years with ezetimibe vs. control¹⁶; the results of this study have yet to be published. As a result of the relatively limited evidence base, cholesterol guidelines are less definitive in treatment algorithms for this particular population, and suggest a shared decision-making approach when considering initiation of statin treatment in individuals ≥75 years of age.¹² This lack of evidence in older individuals is important since clinicians will increasingly face treatment decisions in this quickly-expanding demographic group.^17,18 A particularly challenging subgroup are those older adult patients (≥75 years old) who survive to older age without any traditional cardiovascular risk factors beyond age, sex, or race, but whose predicted risk using existing risk stratification tools remains elevated.

Using data from the National Heart, Lung, and Blood Institute (NHLBI) Pooled Cohorts, containing prospectively measured CVD risk factor profiles and closely adjudicated CVD events, we evaluated the risk of new onset ASCVD among a large cohort of individuals who were free of CVD by age 75. Using these data, we set out to: 1) evaluate the overall unadjusted and risk-adjusted association between LDL-C and CVD events among older adults (≥75 years old) without known ASCVD; and 2) assess the impact of hyperlipidemia on the risk of new ASCVD among older adults with varying risk factor profiles.

METHODS

Data description

Data from the following four cohort studies were obtained from the NHLBI’s Biologic Specimen and Data Repository Information Coordinating Center (BioLINCC) program: Cardiovascular Health Study, Multi-Ethnic Study of Atherosclerosis, Framingham, and Framingham Offspring. We selected 1 exam from each of these 4 studies as the baseline exam for this analysis that would allow at least 5 years of follow-up, while maximizing the number of subjects age 75 or older. In addition, we selected exams from early in the studies prior to widespread statin use in order to minimize the number of subjects on lipid-lowering therapy, which would potentially led to confounding between the association between LDL-C and outcomes. The exams we chose were the Cardiovascular Health Study visit 3 (1992–1993), Multi-Ethnic Study of Atherosclerosis Exam 1 (2000–2002), Framingham Exam 24 (1995–1998), and Framingham Offspring Exam 7 (1998–2001). Patients were included who were free of any CVD by age 75, including heart failure, peripheral arterial disease, coronary heart disease (including angina or prior myocardial infarction), and cerebrovascular disease (including prior stroke or transient ischemic attack).

Variables collected from the baseline exam included age, sex, race, LDL-C levels, lipid-lowering therapy, diabetes and smoking status, body mass index, creatinine clearance (calculated using the Cockcroft-Gault Equation), systolic blood pressure, diastolic blood pressure, and blood pressure-lowering medication. In the selected cohorts, 390 out of 3057 patients ≥75 years old were excluded (12.8%), due to missing 1 or more of the variables used to create the risk factor groups (hyperlipidemia, diabetes, hypertension, or smoking status). Missing variables among patients included in our analysis were rare with 1 out of 2667 observations missing data for LDL-C (in a patient known to be on lipid-lowering treatment) and 4 patients missing data on blood pressure (all of whom were on blood pressure treatment). The primary endpoint was a composite of cardiovascular death, myocardial infarction, and stroke through each subject’s end of follow-up.

Statistical analysis

We assessed baseline characteristics of the population overall and stratified by those with and without hyperlipidemia. Continuous variables are summarized with medians and quartiles and compared with the Wilcoxon test. Categorical variables are presented as frequencies and percentages and compared using chi-square tests. We evaluated the association between LDL-C levels and the risk of CVD events with univariable Cox proportional hazards models, excluding subjects on lipid-lowering therapy. Restricted cubic splines were used to test for departure from linearity in the relationship between LDL-C and the log-hazard of CVD risk. Results assuming both a linear and a non-linear relationship of LDL-C with the log hazard of CVD events are presented. Adjusted hazard ratios (HRs) for LDL-C were derived from multivariable Cox proportional hazards models, including age, sex, race, diabetes, smoking, systolic blood pressure, diastolic blood pressure, and use of blood pressure medication. In sensitivity analysis, analyses were repeated, including adults on lipid-lowering therapy.

We also evaluated the potential impact of the competing risk of non-CV death. First, we evaluated the ratio of non-CV to CV death by LDL-C decile in order to see if there was a difference in the relative causes of mortality by LDL-C decile. Next, we compared the Kaplan-Meier estimates from the Cox models with cumulative incidence estimates calculated using the Fine and Gray approach, which accounts for the competing risk of non-CV death by modelling the sub-distribution hazard of CVD.

Four major risk factors for CVD events were defined for the population: 1) smoking, defined as current smoking status; 2) diabetes; 3) hypertension, defined as systolic blood pressure ≥140, diastolic blood pressure ≥90, or use of blood pressure-lowering medication; and 4) hyperlipidemia, defined as either an LDL-C ≥130 mg/dL or the use of lipid-lowering therapy. An LDL-C cutoff of ≥130 mg/dL was chosen based on clinical conventions and the National Cholesterol Education Program (NCEP) Adult Treatment Panel (ATP) III definition.¹⁹ Patients were then categorized into three risk factor tiers: 1) those without any non-hyperlipidemia risk factors (from current smoking, hypertension, and diabetes); 2) patients with 1 non-hyperlipidemia risk factor (either current smoking, hypertension, or diabetes); and 3) patients with 2 or more non-hyperlipidemia risk factors. Characteristics of adults in each tier were evaluated, stratified by the presence or absence of hyperlipidemia. The probability of having a CVD event during the first 5 years of follow-up was estimated using the Kaplan-Meier method, and survival was compared within each tier between adults with and without hyperlipidemia using the log-rank test. In sensitivity analysis, these comparisons were repeated, excluding adults on lipid-lowering therapy. Analyses using the NHLBI Pooled Cohorts were approved by the Duke University Institutional Review Board (PRO 00051569), and all subjects participating in the individual studies provided informed consent to participate. All analyses were performed using SAS version 9.4 (Cary, NC).

RESULTS

Baseline characteristics

A total of 2667 subjects ≥75 years were identified who were free of CVD across the four cohort studies, including 1199 (45.0%) from the Cardiovascular Health Study, 948 (35.5%) from the Multi-Ethnic Study of Atherosclerosis, 315 (11.8%) from Framingham Original, and 205 (7.7%) from Framingham Offspring. More than half (n=1571, 59%) were female, the median age was 78 years (25^th–75^th percentiles 76–82), median LDL-C was 117 mg/dL (96–140 mg/dL), and 306 (11.5%) were on lipid-lowering therapy. The median follow-up for all participants was 8.3 years, and 25% of subjects had over 10 years of follow-up. The prevalence of major risk factors was 5.4% (n=144) for smoking, 12.5% (n=331) for diabetes, and 67.9% (n=1812) for hypertension. A total of 1166 subjects were considered to have hyperlipidemia, including n=932 with LDL-C ≥130 mg/dL, n=306 on lipid-lowering treatment, and n=72 on lipid-lowering therapy with LDL-C ≥130 mg/dL. Of the 306 subjects on lipid-lowering therapy, 264 were on statin therapy and the others were on non-statin lipid-lowering medications. Characteristics of these adults are presented in Table 1. Among those with hyperlipidemia, the median LDL-C was 143 mg/dL (interquartile range [IQR] 132–158 mg/dL). Among those without hyperlipidemia (n=1501), the median LDL-C was 104 mg/dL (IQR 89–117 mg/dL). Adults with hyperlipidemia were more likely to be female, slightly younger, have higher body mass index, higher average blood pressure, higher total cholesterol, and higher median triglyceride levels (127 mg/dL, IQR 94–168 mg/dL vs. 108 mg/dL, IQR 76–153 mg/dL, p<0.0001).

Table 1.

Characteristics of older adult (≥75 year old) patients in the pooled cohorts by risk group

Characteristic	Overall (N=2667)	Hyperlipidemiaa (N=1166)	No Hyperlipidemiab (N=1501)	p-value
Study cohort				0.0021
CHS	1199 (45.0%)	510 (43.7%)	689 (45.9%)
Framingham Offspring	205 (7.7%)	110 (9.4%)	95 (6.3%)
Framingham Original	315 (11.8%)	118 (10.1%)	197 (13.1%)
MESA	948 (35.5%)	428 (36.7%)	520 (34.6%)
Sex (female)	1571 (58.9%)	767 (65.8%)	804 (53.6%)	<.0001
Age (median, 25th–75th)	78, 76–82	78, 76–81	79, 77–82	0.0001
Age groups				0.0011
75–79	1760 (66.0%)	809 (69.4%)	951 (63.4%)
80 or more	907 (34.0%)	357 (30.6%)	550 (36.6%)
Race				0.7827
White	2045 (76.7%)	897 (76.9%)	1148 (76.5%)
African-American	296 (11.1%)	132 (11.3%)	164 (10.9%)
Other	326 (12.2%)	137 (11.7%)	189 (12.6%)
BMI (median, 25th–75th)	26.0, 23.3–29.0	26.2, 23.9–29.2	25.8, 23.0–28.8	0.0004
SBP (median, 25th–75th)	137, 122–152	138, 124–152	136, 121–152	0.0485
DBP (median, 25th–75th)	69, 63–76	69, 63–77	69, 62–76	0.1240
SBP ≥140	1194 (44.8%)	542 (46.5%)	652 (43.6%)	0.1315
DBP ≥90	98 (3.7%)	39 (3.4%)	59 (4.0%)	0.4127
SBP/DBP ≥140/90	1201 (45.0%)	544 (46.7%)	657 (43.8%)	0.0183
On BP meds	1247 (46.8%)	549 (47.1%)	698 (46.5%)	0.7652
Total cholesterol (median, 25th–75th)	197, 173–222	224, 205–245	182, 165–197	<.0001
HDL-C (median, 25th–75th)	52, 43–62	52, 44–60	53, 42–64	0.2028
Triglycerides (median, 25th–75th)	117, 83–160	127, 94–168	108, 76–153	<.0001
LDL-C (median, 25th–75th)	117, 96–140	143, 132–158	104, 89–117	<.0001
LDL-C ≥130	932 (34.9%)	932 (79.9%)	0 (0.0%)	<.0001
On lipid-lowering meds	306 (11.5%)	306 (26.2%)	0 (0.0%)	<.0001
Diabetes	331 (12.4%)	140 (12.0%)	191 (12.7%)	0.5769
Current smoker	144 (5.4%)	59 (5.1%)	85 (5.7%)	0.4944
eGFR (median, 25th–75th)	53.1, 43.6–64.0	53.2, 43.9–63.1	53.1, 43.5–64.7	0.4171
On Aspirin	498 (18.7%)	242 (20.8%)	256 (17.1%)	0.0150

Abbreviations: BMI, body mass index; BP, blood pressure; CHS, Cardiovascular Health Study; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; MESA, Multi-Ethnic Study of Atherosclerosis; SBP, systolic blood pressure

^aThe hyperlipidemia group includes individuals on lipid-lowering medications and/or with LDL-C≥130 mg/dL.

^bThe no hyperlipidemia column includes participants not on lipid-lowering medications and with LDL-C<130 mg/dL.

^cIncludes Asian, Hispanic and other races/ethnicities.

Association between LDL-C and ASCVD risk

In a univariable analysis, there was no association between LDL-C and CVD events in adults ≥75 years (HR per 10-unit increase=1.009, 95% confidence interval [CI] 0.986–1.032, p=0.46; Figure 1). Departure from linearity between LDL-C and the log-hazard of CVD events in the unadjusted model was mild (p-value for non-linearity=0.09); however, even when modeled as a non-linear association, the relationship between LDL-C and CVD events was not statistically significant (p=0.15). The lack of association between LDL-C and CVD events remained non-significant after adjusting for other baseline CVD risk factors (adjusted HR 1.022, 95% CI 0.998–1.046, p=0.07). Interactions for age (<80/80+) and race were assessed and did not change the results. In a sensitivity analysis that included adults on lipid-lowering therapy (n=306), no association was seen between LDL-C and CVD event risk on unadjusted analyses (unadjusted HR 1.013, 95% CI 0.991–1.036, p=0.25), and a borderline association after adjusting for other CVD risk factors (adjusted HR 1.024, 95% CI 1.001–1.047, p=0.04) (Supplementary Figure S1). There was a slightly higher ratio of non-cardiovascular death events to cardiovascular deaths in the lowest three LDL-C deciles (Supplementary Table S1); however both Kaplan-Merier and cumulative incidence estimates yielded similar findings, with a lack of association between LDL-C and CVD events (Supplementary Table S2).

Figure 1. LDL-C and ASCVD risk among untreated older adults (≥75 years).

This figure demonstrates the results from the univariable evaluation of association between LDL-C and ASCVD risk with LDL-C modeled as a linear variable (HR per 10 unit increase was 1.009 [0.986–1.032], p=0.46) and using restricted cubic splines. In multivariable modeling the HR was 1.022 (95% CI 0.998–1.046, p=0.07), with a p-value for non-linearity=0.40. The model for non-linear LDL does not significantly improve the fit and there was no significant evidence of a non-linear relationship between LDL-C and ASCVD in the unadjusted model (p=0.15); 306 observations on lipid-lowering medications were excluded from this analysis. Abbreviations: ASCVD, atherosclerotic cardiovascular disease; CI, confidence interval; HR, hazard ratio; LDL-C, low-density lipoprotein cholesterol

Association between hyperlipidemia and ASCVD risk in the presence or absence of other risk factors

Overall, we identified 737 subjects without any non-lipid risk factors, 1586 with 1 non-lipid risk factor, and 344 with 2 or more non-lipid risk factors. The prevalence of hyperlipidemia in these 3 groups was 41.8%, 45.1%, and 41.6% respectively. Baseline characteristics by risk factor tier are presented in Supplementary Table S3. Figures 2A, 2B, and 2C show Kaplan-Meier event rates for adults in each risk factor tier, stratified by the presence or absence of hyperlipidemia. There was no difference in the observed 5-year event rate for adults with and without hyperlipidemia in any of the three tiers. Specifically, among those without non-lipid risk factors, 5-year ASCVD risk was 5.8% (95% CI 3.6–9.1%) for individuals with hyperlipidemia vs. 7.0% (95% CI 4.9–10.0%) for individuals without hyperlipidemia (p=0.25, Figure 2A). For those with 1 non-lipid risk factor, the risk was similar for individuals with hyperlipidemia (Kaplan-Meier rate=12.8%, 95% CI 10.5–15.5%) as those without (Kaplan-Meier rate 15.0, 95% CI 12.7–−17.8%, p-difference=0.44, Figure 2B). For those with 2 or more non–lipid-related risk factors, those with hyperlipidemia (Kaplan-Meier rate 21.9%, 95% CI 15.8–30.0%) had no difference in event rate compared with those without (Kaplan-Meier rate 24.0%, 95% CI 18.5–31.0%, p-difference=0.59, Figure 2C). The lack of difference between event rates in those with and without hyperlipidemia persisted even after excluding those on lipid-lowering therapy (Supplementary Table S4).

Figure 2. Association between hyperlipidemia and ASCVD risk in the presence or absence of other risk factors.

This figure demonstrates the observed ASCVD risk of older adults (≥75 years old) with: A) no traditional risk factors (no current smoking, diabetes, hypertension, or hyperlipidemia) vs. individuals with no risk factors other than hyperlipidemia (defined as LDL-C ≥130 mg/dL or on lipid-lowering therapy); B) 1 traditional risk factor (from current smoking, diabetes, or hypertension) vs. individuals with 1 risk factor plus hyperlipidemia (defined as LDL-C ≥130 mg/dL or on lipid-lowering therapy); and C) 2 or more traditional risk factors (from current smoking, diabetes, or hypertension) vs. individuals with 2 or more risk factors plus hyperlipidemia (defined as LDL-C ≥130 mg/dL or on lipid-lowering therapy). Abbreviations: ASCVD, atherosclerotic cardiovascular disease; LDL-C, low-density lipoprotein cholesterol; HLD, hyperlipidemia

DISCUSSION

Observational data have consistently shown that hyperlipidemia is a risk factor for CVD among young and middle-aged adults. Nevertheless, it is unclear if this risk persists among those who survive to older ages free of ASCVD. Our study of 2667 older adults (≥75 years old) without known ASCVD found no association between LDL-C levels and ASCVD risk. These results were robust to the form with which LDL-C was modeled and whether other potential confounders were included. At a population level, when subjects with and without hyperlipidemia were compared within three risk factor tiers (0, 1, or 2+ risk factors from smoking, diabetes, and hypertension), cardiovascular event rates were similar. Combined, these data suggest that among older adults without CVD, increases in LDL-C do not inflate ASCVD risk.

The association between elevated LDL-C and future cardiovascular events is well-established in younger populations^12,20–22; however, literature is mixed for older adults, with some studies identifying an association between elevated LDL-C and cardiovascular risk (even among individuals ≥75 years old),^4,7,8 while other studies have demonstrated no association or an inverse association.^6,9–11 One meta-analysis of 61 prospective observational studies and almost 900,000 subjects demonstrated a positive association between total cholesterol and ischemic heart disease mortality even at older ages, but included a highly heterogeneous patient population whose only inclusion/exclusion criteria was that the patient not have a history of stroke or heart disease; have blood pressure, cholesterol, age, and sex data available at enrollment; have more than 5000 person-years of follow-up; and have cause/date of death routinely sought. The study did not account for the use of lipid lowering therapies. A separate meta-analysis of 19 studies with less expansive inclusion criteria found no association between LDL-C and mortality among patients older than 60 years,¹¹ but also did not account for baseline statin therapy. Neither of these studies captured composite cardiovascular disease, including cardiovascular events such as myocardial infarction and stroke. Our study contributes important findings to the existing literature by accounting for the presence or absence of lipid-lowering therapy in a large population of older U.S. adults (≥75 years) with close follow-up and adjudication of cardiovascular events, in addition to cardiovascular mortality, and a dedicated consideration of the important competing risk of non-cardiovascular death. Clinicians routinely consider LDL-C as a key risk factor for future cardiovascular events in older adults; the lack of association between LDL-C and future risk of CVD identified in our study contradicts this common practice and emphasizes the need to consider other cardiovascular risk factors in this population while developing specific risk stratification tools in older adults.

While LDL-C levels may not correlate with the risk of CVD in an older, primary prevention population, our findings should not be interpreted to suggest there is no benefit to lipid-lowering therapy in older adults. Rather, our findings indicate those at high-risk for CVD based on non–lipid-related risk factors such as hypertension, diabetes, and smoking should be targeted for lipid-lowering therapy regardless of starting LDL-C levels. The effectiveness of statin therapy to lower CVD risk is supported by a wealth of evidence demonstrating the benefit of LDL-C lowering with statin and non-statin lipid-lowering therapies,^{1,2,12,23–25} with no lower LDL-C limit for ASCVD risk reduction,^26–28 and a benefit proportional to absolute risk and LDL-C reduction.³ Yet to date, there are limited clinical trial data evaluating the effectiveness of lipid-lowering therapy in older adults without established ASCVD.^14,15,29 Our finding of a lack of association between LDL-C and CVD risk in older adults does not indicate whether statin therapy can be useful for lowering CVD risk. The benefit of statin therapy has been shown across a wide range of starting LDL-C levels. Rather, our finding suggests that other risk factors for CVD beyond LDL-C should be used to identify those at highest risk of CVD in order to guide statin therapy. However, the question of the effect of statin therapy in older adults remains unanswered. A recent CTT meta-analysis evaluating the efficacy of statin therapy for primary prevention found no significant risk reductions in individuals >70 years old.²³ Taken together, these findings highlight the need for dedicated randomized trial evidence for lipid-lowering therapy in the older adult population. The Statins in Reducing Events in the Elderly (STAREE) trial, which has been ongoing in Australia since 2015, is investigating the primary prevention impact of atorvastatin vs. placebo in ≥70 year old individuals independently living in the community. In addition, a randomized clinical trial investigating the benefit of lipid-lowering therapy for the primary prevention of CVD among older adults is currently proposed by the NHLBI and should provide tremendous insight on this issue.

A lack of association between LDL-C and cardiovascular events in younger (but not older) adults may be explained by a number of biological factors. First, younger adults with elevations in LDL-C have a longer potential exposure duration to elevations in LDL-C compared with older adults who may have developed dyslipidemia later in life, leading to increased potential risk.³⁰ Given the strong correlation between duration of exposure to LDL-C and cardiovascular risk, many of those adults with earlier onset elevations in LDL-C are likely to have developed ASCVD by the age of 75 and, therefore, would not have been included in the analysis. Second, not all phenotypes of elevations in LDL-C are equally atherogenic; when LDL-C mass is elevated in the absence of a corresponding elevation in particle number, risk is attenuated.³¹ Adults with elevated LDL-C who have not yet developed ASCVD may possibly have low particle numbers and, consequently, relatively less atherogenic particles. Similarly, while risk factors for CVD are well-described, it is not yet understood why some individuals develop disease while others do not. By selecting for adults without ASCVD by age 75, we may have created a selection bias towards those adults with other, to-date, unknown factors protecting them from the risk of CVD conferred by lipid elevations.

Accounting for competing risk of death in an older adult population is critical to the interpretation of the association between LDL-C and CVD events. An association between low LDL-C and non-cardiovascular mortality, including cancer mortality, has been previously reported in the literature^32,33. While we similarly observed higher non-cardiovascular mortality at low LDL-C levels, our Cox proportional hazard model accounted for competing risks by modelling the cause-specific hazard of CVD. An alternative to the Cox model would have been the Fine and Gray model which is based on the sub-distribution hazard of CVD. However, models based on the sub-distribution hazard are more appropriate when predicting an individual risk while cause-specific hazard models are better suited for studying the etiology of diseases³⁴. Nonetheless, we evaluated the two approaches and found that both Kaplan-Meier and cumulative incidence estimates provide approximately the same estimates. This reassures us that even when competing risk is accounted for, there was no significant association between LDL-C and CVD events.

Our study has some limitations. First, we evaluated key traditional cardiovascular risk factors: age, race, sex, current smoking, diabetes, hypertension, and hyperlipidemia; however, there are a multitude of other risk factors that may impact LDL-C, future cardiovascular risk and treatment decisions, including other lifestyle factors, malignancy, inflammatory disorders, premature menopause, ethnicity, and socioeconomic and geographic factors. Second, in order to obtain a sample of adults with minimal contamination by the use of lipid-lowering therapy, we relied on relatively older data. Temporal trends in treatment patterns and ASCVD-free survival may alter the association between LDL-C and ASCVD risk over time as more adults reach age 75 without ASCVD. Third, we did not evaluate the use of lipid lowering drugs or LDL-C longitudinally, which prevents us from assessing the potential impact of changes in medication use and LDL-C over time on cardiovascular risk. Finally, our analyses were somewhat limited by the modest number of older adults ≥75 years old in the pooled cohorts, which prevents the detection of smaller differences between groups.

In conclusion, we observed no association between LDL-C levels and CVD risk in adults ≥75 years, and hyperlipidemia did not add incremental risk beyond other traditional cardiovascular risk factors across the risk spectrum. A patient’s absolute risk of CVD, and the presence of other non–lipid-related risk factors, should be used to guide treatment decisions for lipid-lowering therapy.