Study design
The present study was carried out within the framework of the PREDIMED study, a large randomized, multicenter, parallel-group, clinical trial aiming to assess the effects of MedDiet on the primary prevention of CVD in a cohort of older individuals at high cardiovascular risk. Participants were aged between 55 and 80 years and had no CVD at enrollment, but they were at high risk because of the presence of type 2 diabetes or at least three of the following risk factors: current smoking, hypertension, hypercholesterolemia, low high-density lipoprotein (HDL)-cholesterol, overweight or obesity, and family history of premature CVD. Exclusion criteria included any severe chronic illness, drug or alcohol addiction, or allergy or intolerance to olive oil or nuts, two key supplemental foods. In the main study, participants were randomly assigned to three intervention groups: a MedDiet supplemented with virgin olive oil, a MedDiet supplemented with mixed nuts, or a low-fat diet according to the American Heart Association guidelines (control group). The trial is registered at http://www.controlled-trials.com as ISRCTN35739639 and the study protocol and results of the primary outcome have been published elsewhere [4]. The PREDIMED trial was conducted according to the Helsinki Declaration, and the institutional review boards of all recruiting centers approved the study protocol (for the Reus center, the protocol was approved by Hospital Universitari Sant Joan de Reus Ethical Committee). Participants agreed and gave their written informed consent to authorize the use of biological samples for biochemical measurements and genetic studies.
The present cross-sectional and longitudinal analyses were conducted on 196 participants recruited in the PREDIMED-Reus (Spain) center with full food frequency questionnaire (FFQ) data and plasma samples available at baseline and 1 year of follow-up.
Dietary assessment
Trained dietitians assessed dietary intake in face-to-face interviews at baseline using a validated semi-quantitative 137-item FFQ [10]. For each item, the portion size was established, and nine consumption frequencies were available, ranging from “never or rarely” to “ ≥ 6 times/day”. Energy and nutrient intakes were obtained using data from Spanish food composition tables [11].
Data on self-reported nut consumption were derived from the FFQ, which included an item on the consumption of almonds, peanuts, hazelnuts, pistachios, and pine nuts, and another specific question on the consumption of walnuts. For the present analysis, 28 g of nuts was considered a serving. The number of reported servings was converted into grams per day. The Pearson correlation coefficients for reproducibility and validity of the FFQ regarding nut consumption were 0.66 and 0.38, respectively, and intraclass correlation coefficients for the same measurements in a similar population to the PREDIMED participants were 0.80 and 0.55, respectively [10].
Covariates
Information about sociodemographic and lifestyle variables, including smoking status, medical conditions, family history of the disease, and medication use, were collected at baseline. Physical activity was estimated with the validated Spanish version of the Minnesota Leisure Time Physical Activity Questionnaire [12]. Trained staff measured height and bodyweight without shoes and wearing light clothing to the nearest 0.5 cm for height and 0.1 kg for bodyweight using a wall-mounted stadiometer and calibrated scales, respectively.
Lipoprotein and metabolite profiling
Fasting blood samples were collected at baseline and 1-year visits. EDTA plasma was obtained and aliquots were stored at – 80 °C until metabolomic analysis. NMR spectra were acquired on a Vantera® Clinical Analyzer, a 400 MHz NMR instrument, from EDTA plasma samples as described for the NMR LipoProfile® test (Labcorp, Morrisville, NC) [13, 14]. The LP4 deconvolution algorithm was used to report lipoprotein particle concentrations and sizes, as well as concentrations of metabolites such as total branched-chain amino acids, valine, leucine, isoleucine, alanine, glucose, citrate, glycine, total ketone bodies, β-hydroxybutyrate, acetoacetate, acetone [15]. The diameters of the various lipoprotein classes and subclasses are total triglyceride-rich lipoprotein particles (VLDL-P) (24–240 nm), very large VLDL-P (90–240 nm), large VLDL-P (50–89 nm), medium VLDL-P (37–49 nm), small VLDL-P (30–36 nm), very small VLDL-P (24–29 nm), total low-density lipoprotein particles (LDL-P) (19–23 nm), large LDL-P (21.5–23 nm), medium LDL-P (20.5–21.4 nm), small LDL-P (19–20.4 nm), total high-density lipoprotein particles (HDL-P) (7.4–12.0 nm), large HDL-P (10.3–12.0 nm), medium HDL-P (8.7–9.5 nm), and small HDL-P (7.4–7.8 nm). The peak diameters for the largest (H7) to the smallest (H1) of the HDL subspecies are 12.0, 10.8, 10.3, 9.5, 8.7, 7.8, and 7.4 nm. Mean VLDL, LDL, and HDL particle sizes are weighted averages derived from the sum of the diameters of each of the subclasses multiplied by the relative mass percentage. Linear regression against serum lipids measured chemically in a healthy study population (n = 698) provided the conversion factors to generate NMR-derived concentrations of total cholesterol (TC), triglycerides (TG), VLDL-TG, VLDL-C, LDL-C, and HDL-C. NMR-derived concentrations of these parameters are highly correlated with those measured by standard chemistry methods. Details regarding the performance of the assays that quantify branched-chain amino acids (BCAA), citrate, and ketone bodies have been reported [16,17,18]. Development of the Lipoprotein Insulin Resistance Index (LP-IR) (0–100; least to most insulin resistant), the Diabetes Risk Index (DRI) (1–100; the lowest to the highest risk of type 2 diabetes), and GlycA, a composite measure of systemic inflammation, as well as their analytical and clinical validation, have been published previously [19,20,21].
Statistical analyses
Participantsʼ baseline characteristics are described as means ± SD for quantitative traits and percentages for categorical traits. Nut consumption at baseline and 1-year changes were adjusted for total energy intake using the residual method [22]. Nut consumption was categorized into tertiles according to total nuts, walnuts, or non-walnut nuts at baseline and 1-year changes.
Baseline values and 1-year changes in individual lipoprotein, lipid, apolipoprotein, amino acid, ketone bodies, and other molecules were normalized and scaled using Blom's rank-based inverse normal transformation to improve normality [23].
We assessed differences in lipoprotein values between tertiles of nut consumption (total nuts, walnuts, and non-walnut nuts) at baseline using ANCOVA models adjusted by age, gender, body mass index (kg/m2), smoking status (ever smoker/never smoker), physical activity (met/day), diabetes (yes/no), dyslipidemia (yes/no), hypertension (yes/no), and statin treatment (yes/no). Data are presented as means and 95% confidence intervals (CI). We also assessed differences in 1-year changes in lipoprotein values between tertiles of nut consumption using ANCOVA models additionally adjusted by the baseline lipid value, baseline nuts consumption, and intervention group (MedDiet + EVOO, MedDiet + Nuts, Low-fat diet). The Tukey test was used to perform multiple comparisons between tertiles. We repeated the same analyses with other molecules including apolipoprotein, amino acids, and ketone bodies.
The assumptions of the ANCOVA models were assessed using visual or quantitative methods. All graphs and tests (Shapiro–Wilk test and Levene’s tests) yielded models that met the criteria for the independence of observations, homogeneity of variance (all Levene’s test P values > 0.05), and normality of residuals (all Shapiro–Wilk test P values > 0.05).
P values < 0.05 were considered statistically significant for these analyses. All statistical analyses were performed with the R software v3.6.1 (www.r-project.org) (R Development Core Team, 2012).