This investigation used data from the Atherosclerosis Risk in Communities (ARIC), Multiethnic Study of Atherosclerosis (MESA) and Coronary Artery Risk Development in Young Adults (CARDIA) studies. Each prospective cohort study consisted mostly of healthy adults without cardiovascular disease recruited from the general population, and collected similar key variables including blood cholesterol and ASCVD events occurring during long-term follow-up. All participants provided informed consent for study participation at each study visit. Institutional review boards at all participating sites have previously approved the 3 studies.

The current observational analysis used pooled, deidentified individual-level data from the 3 cohorts. The baseline visit was determined based on the time of the first blood cholesterol measurement in each cohort: Visit 4 in ARIC (1996-1998), Visit 1 in MESA (2000-2002) and Visit 1 in CARDIA (1985). The end of follow-up was defined as December 31, 2016 for ARIC, December 31, 2023 for MESA and December 31, 2011 for CARDIA.

The study population included all adults free of ASCVD at baseline. The age ranges at the baseline exam in each cohort were 45-65 years in ARIC, 45-84 years in MESA and 18-30 years in CARDIA. Adults with existing ASCVD ( n = 1,292), missing values for lipids measurements ( n = 2,213) and lack of available follow-up times ( n = 6) were excluded. Since lipid-lowering therapies are prescribed based on individual clinical characteristics and at varying time points, adults taking lipid-lowering therapy at baseline ( n = 1,148) were also excluded to avoid biases in observed associations between lipid measures and long-term events; data on lipid-lowering therapy after the baseline visit were not available. Pooled data were previously constituted to include only White and Black adults. Data for Hispanics ( n = 1,496) and Chinese ( n = 803) adults in MESA, and non-White or non-Black adults in ARIC ( n = 69) were therefore not available for analysis. The final study cohort included 16,384 participants without ASCVD and not on lipid-lowering therapy at their baseline visit.

In the pooled dataset, lipid measurements were only uniformly available at the baseline visit. Measurements of total cholesterol (TC) and triglycerides (TG) in ARIC were based on enzymatic assays and HDL-C was measured using dextran-magnesium precipitation. ApoB in ARIC was determined using World Health Organization/International Federation of Clinical Chemistry (WHO/IFCC) standardized reference materials. In MESA, fasting blood draws were collected and stored at −70 °C, and lipids were measured using eDTA plasma at a central laboratory within 2 weeks of collection. Cholesterol oxidase methods were used to measure TC and TG were measured using TG GB on a Roche COBAS FARA centrifugal analyzer (Roche Diagnostics, Indianapolis, IN, USA). HDL-C in MESA was measured using cholesterol oxidase methods after precipitation of non-HDL-C with magnesium/dextran (Roche Diagnostics). ApoB in MESA was measured with the Tina-quant apoB ver.2 immunoassay on a Roche Modular P analyzer (Roche Diagnostics). In CARDIA, TC and HDL-C were measuring using fasting blood draws that were separated into plasma and frozen at −70 °C prior to central laboratory analysis. ApoB in CARDIA was measured using an enzyme-linked immunosorbent assay. For all analyses, the Martin-Hopkins equation was used to calculate LDL-C based on non-HDL-C and TG. ^,

The main exposures for this study were blood cholesterol measures with participants stratified into categories of baseline serum levels. Low-normal serum levels of LDL-C, non-HDL-C and apoB were defined as <70 mg/dL, <100 mg/dL and <80 mg/dL, respectively. Participants with high-normal levels (70-99 mg/dL for LDL-C, 100-129 mg/dL for non-HDL-C, and 80-89 mg/dL for apoB) were used as the reference group. Participants with elevated levels of LDL-C (≥100 mg/dL, non-HDL-C (≥130 mg/dL) and apoB (≥90 mg/dL) were also evaluated and compared to the reference group. These categories were chosen a priori based on cutoffs proposed in previous guidelines, ^,,, even though population-level percentiles can vary among the different measures. ^,,,

Covariates included demographics (age, sex, race) and other cardiovascular risk factors which were obtained from medical history, physical examination and laboratory data at the baseline visit. These included body mass index, smoking status (never, former or current), systolic blood pressure and diabetes mellitus. In ARIC, 2 blood pressure measurements were taken and the mean of both measurements was used for analyses. In MESA and CARDIA, 3 blood pressure measurements were taken and the mean of the second and third was used for analyses. Diabetes mellitus was defined as a fasting serum glucose ≥126 mg/dL, a nonfasting glucose ≥200 mg/dL, a self-reported physician diagnosis of diabetes, or use of a hypoglycemic medication.

The primary outcome of this study was incident acute ASCVD events, defined as a composite of definite or probable myocardial infarction, stroke or coronary death. Individual ASCVD events were defined similarly across the 3 cohorts and were determined from annual telephone calls, medical records, hospital discharge summaries and ICD codes, and/or death certificates, as has been previously described. ^,,,,, In ARIC, study investigators performed continuous monitoring for all cardiovascular disease-related hospitalizations and deaths among cohort participants, and directly adjudicated all ASCVD events including coronary deaths. Monitoring in ARIC was performed using annual interviews or telephone calls with participants or next of kin, review and abstraction of hospital medical records, and questionnaires completed by physicians, medical examiners, and/or coroners for participants who died. In MESA, outcomes data were collected based on telephone calls at 9-12 month intervals, patient information at MESA visits, and medical records including hospitalization records collected in follow up. Two physicians were blinded to participant data and independently adjudicated all ASCVD events. For suspected coronary deaths, a committee of MESA physicians reviewed information from death certificates and available medical records, and obtained additional information from interviews with next of kin and physician questionnaires for out of hospital deaths. In CARDIA, participants were contacted annually by telephone and medical records were requested for those who were hospitalized or received an outpatient revascularization procedure. Two physician members on the CARDIA endpoints surveillance and adjudication committee independently adjudicated all events, and the full committee made final decisions for events with disagreement. For deaths, death certificates, hospital records, autopsy reports, and physician questionnaires were reviewed. When insufficient information was available to determine the cause of death, interviews were also conducted with family members, friends and/or other individuals who were present when the participant died.

Baseline characteristics of study participants were described with stratification by serum atherogenic lipoproteins at baseline (eg, <70 mg/dL, 70-99 mg/dL, 100-129 mg/dL, and ≥130 mg/dL for LDL-C). Kaplan–Meier estimates of cumulative event-free survival were used to assess long-term incidence of ASCVD events, and log-rank tests were used for comparisons. In prospective analyses Cox proportional hazards models were used to assess the independent association of baseline blood cholesterol measures with incident ASCVD events. LDL-C, non-HDL-C and apoB were evaluated separately with stratification into 4 groups, with the high-normal group used as the reference. Three adjusted models were used. Model 1 was adjusted for baseline age, sex and race. Model 2 was additionally adjusted for baseline body mass index, hypertension, systolic blood pressure, smoking and diabetes, and Model 3 adjusted for all variables in Model 2 in addition to TG.

In secondary analyses, adjusted analyses were performed with stratification by baseline age of <55 and ≥55 years, since most adults over 55 years of age are recommended for statins based on current guidelines. Analyses were also performed based on sex and study cohort. Adjusted analyses were also performed by including a separate group with very low baseline blood cholesterol (LDL-C < 55 mg/dL, non-HDL-C < 85 mg/dL or apoB <60 mg/dL). In a smaller subgroup of participants with baseline high-sensitivity C-reactive protein (hsCRP) available ( n = 11,474), associations were determined by hsCRP level (<2 or ≥2 mg/dL). Lastly, given that cutoffs for LDL-C, non-HDL-C, and apoB represent different percentiles across the population, we performed analyses using lipoprotein quartiles to ensure an equal number of participants in each group across all measures.