Research design and methods
Study population
The ARIC Study is an ongoing community-based cohort of 15 792 men and women (mostly whites and African-Americans) aged 45–64 years at baseline (1987–1989) who were probabilistically selected from four US communities (Forsyth County, North Carolina; Jackson, Mississippi; suburbs of Minneapolis, Minnesota and Washington County).10 Our study sample originated from 1114 ARIC participants invited into a nested case–control study during ARIC visit 2, which occurred from 1990 to 1993.1 ,11 This substudy, composed of cases with subclinical carotid atherosclerosis and controls with minimal or no carotid artery atherosclerosis, was designed to examine the relationship between PPL and subclinical atherosclerosis, among participants without prevalent CVD.1 ,11
Fat-tolerance test
After a 12-hour period of fasting and avoidance of exercise, fasting blood specimens were collected. A liquid test meal was then administered, which consisted of heavy whipping cream, ice cream, safflower oil, chocolate syrup and powdered protein (Promod, Ross Laboratories) and contained 1265 kcal, 32 g protein, 48 g carbohydrate, 105 g fat (52 g saturated) and 300 mg cholesterol, in addition to 100 000 IU vitamin A (Aquasol, Armour Pharmaceutical Co.) for each 2 m2 of body surface area. As a safeguard against potential lactose intolerance, Lactaid was added to the meal. Participants were allotted 15 min to consume the test meal and instructed to ingest only water, unsweetened coffee or tea or sugarless soft drinks for 8 hours thereafter. Blood specimens were drawn 3.5 and 8 hours following test meal consumption. The number of postprandial blood collections (ie, the difference between two and four measurements) has shown minimal influence on the classification of postprandial response.2 Hence, measurements of PPL variables taken post-fasting and then 3.5 and 8 hours after consumption of the test meal should be suitable to assess postprandial response to a fatty meal over an 8-hour time period.
Blood specimens were centrifuged for 20 min (1500 g, at 4°C) to separate plasma. Plasma samples were shipped on dry ice and stored in liquid nitrogen at the central laboratory until analysis within a few days. Details of this procedure have been described elsewhere.1 ,11
PPL assays
Variation in CVD risk by subclasses of PPL-related variables was considered by using four separate measures of postprandial markers (each assessed at visit 2): postprandial triglycerides (TG), TG-rich lipoprotein triglycerides (TRL-TG), retinyl palmitate (RP) and apolipoprotein B48 to apolipoprotein B100 ratio (apoBR). TG and TRL-TG appear in the ‘top fraction’ and are markers of very low-density lipoproteins, whereas measures of RP and apoBR are surrogates for chylomicrons. In this study, TG levels were measured enzymatically, TRL-TG samples were isolated by ultracentrifugation,12 plasma RP levels were determined using high-pressure liquid chromatography13 and apoBR values were ascertained by sodium dodecyl sulfate electrophoresis.14
To yield a measure of postprandial change, the incremental area under the curve (iAUC) defined by each PPL parameter across the available time points (0, 3.5 and 8 hours after the test meal) was calculated, with negative results set to 0.1 For the purposes of this study, the iAUC can be interpreted as the two-dimensional postprandial response over 8 hours. The iAUC measures were transformed into weighted SD units to reflect the sampling weights of the overall ARIC sample from which case–control participants were drawn. Weighted SD units can be interpreted in the same manner as conventional SD units.
Other covariates
HDL-C and low-density lipoprotein cholesterol (LDL-C) were measured at visit 2.15 ,16 Trained ARIC personnel used a standardised protocol to measure height, weight, former as well as current cigarette smoking status and diastolic and systolic blood pressure at visit 2.17 Prevalent hypertension was indicated by a systolic blood pressure ≥140 mm Hg, a diastolic blood pressure ≥90 mm Hg and/or self-reported use of antihypertensive medication. The use of lipid-lowering medications was self-reported by the patient and verified by ARIC personnel (checking of medications brought by the patient) at visit 2. Fasting blood glucose measures were also taken at visit 2.
Severity of atherosclerosis was assessed using B-mode ultrasound as in the aforementioned nested case–control study during visit 2.1 ,11 In that study, white participants with IMT values above the 95th centile of the ARIC distribution were sampled as cases, while the 90th centile was applied to obtain a sufficient number of cases among black participants. Controls in race/ethnicity groups were below the 75th centile of IMT in all artery segments evaluated, as IMT values greater than the 75th centile are considered abnormal and indicative of increased CVD risk.18 Cases from that study were classified as having a high severity of atherosclerosis in this study, and controls as having a low severity of atherosclerosis.
Ascertainment of CVD
Incident CHD events included definite or probable myocardial infarction (MI), ‘silent’ MI per centrally interpreted ECG, definite CHD death or coronary revascularisation. This information was obtained through annual contact of participants, yearly discharge lists from local hospitals and ongoing survey of death certificates.17 Incident (ischaemic) stroke was identified through hospital discharge codes and deaths attributed to stroke.19 Physician reviewers adjudicated all possible CHD and stroke events and classified them as definite, probable or non-events based on information abstracted from medical records.19 Cohort follow-up of incident CHD and stroke events through 31 December 2012 was included in these analyses.
Statistical analyses
Plots of residuals depicted a relatively linear relationship between each PPL parameter (as the iAUC-transformed weighted SD measure) and incident CVD (CHD or stroke). For each PPL parameter, we therefore estimated crude, age-sex adjusted and fully adjusted associations between an increase of 1-unit in the iAUC-transformed weighted SD measure and incident CVD events (CHD individually, stroke individually, and CHD or stroke as a merged outcome). Cox proportional hazard models20 were used to quantify estimates of associations on the relative scale via HRs, and incident rate differences (IRDs) from Poisson models21 were used as estimates on the absolute scale.
Since knowledge of the biological mechanism involved in postprandial change and CVD development is not clear, we applied stepwise selection (p<0.15 for inclusion and exclusion)—as opposed to a priori adjustment—to guide in the choice of covariates for fully adjusted models. The following variables, each at visit 2, were considered in such models: age, sex, race (white or African-American), study center location, BMI (≥30 kg/m2 or<30 kg/m2), atherosclerosis severity (high or low), HDL-C, LDL-C, fasting blood glucose, current cigarette smoking, former cigarette smoking, prevalent hypertension and use of lipid-lowering medications.
As per findings from the ARIC case–control study of PPL and atherosclerosis,1 we also estimated associations of our PPL measures with incident CVD by dichotomies of race, BMI and atherosclerosis severity. Comparison of effect measures across these subgroups of low versus high cardiometabolic risk may provide insights into the temporality and etiology of potential mechanisms. For instance, evidence of an independent association between excessive PPL and incident CVD in participants with low but not high IMT would suggest that PPL exerts its role in the pathophysiology of CVD prior to or during the earlier stages of atherosclerosis. In this study, all statistical analyses were performed using SAS V.9.3 (SAS Corp, Cary, North Carolina, USA).