Frequency of nut consumption and mortality risk in the PREDIMED nutrition intervention trial

The present study was conducted within the framework of the PREDIMED trial, the design of which has been described in detail elsewhere [14]. Briefly, the PREDIMED study is a large, multicenter, parallel-group, randomized and controlled clinical trial for the primary prevention of cardiovascular disease (CVD) (http://​www.​predimed.​es and http://​www.​predimed.​org). The main results of the trial on the primary endpoint have been recently published [15]. We assigned 7,447 older participants (men aged 55 to 80 years and women 60 to 80 years) to 1 of 3 interventions: a MedDiet enriched with extra-virgin olive oil (EVOO), a MedDiet supplemented with mixed nuts, or advice on a low-fat diet (control diet). Participants had no CVD at enrollment but they were at high cardiovascular 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 were the presence of severe medical condition that may impair the ability of the person to participate in a nutrition intervention study (for example, digestive disease with fat intolerance, advanced malignancy, or major neurological, psychiatric or endocrine disease), immunodeficiency or HIV positive status, alcohol or drug abuse, body mass index (BMI) ≥40 kg/m², and allergy or intolerance to olive oil or nuts [16].

The primary endpoint of the main trial is a combination of several cardiovascular events (myocardial infarction, stroke or cardiovascular death). The present study was conducted as an observational cohort using baseline consumption of nuts as the exposure. The outcomes were: (1) total mortality, (2) only cardiovascular mortality, and (3) only cancer mortality. All participants provided written informed consent according to a protocol approved by the institutional review boards of the recruiting centers (Comité de Ética e Investigación Clínica (CEIC) Hospital Universitari Sant Joan de Reus, CEIC Universidad de Navarra, CEIC Hospital Clínic de Barcelona, Comité de Ética Universidad de Valencia, CEIC-Parc de Salut Mar, CEIC Hospital Universitario Araba, CEIS del distrito Sanitario Atención Primária Sevilla, IDIAP Jordi Gol, CEIC Complejo Hospitalario Materno-Insular, CEIC Facultad Medicina Universidad de Málaga, CEIC Illes Balears, and CEIC Hospital Universitari Bellvitge).

At baseline trained dietitians completed a 137-item semiquantitative food frequency questionnaire in a face-to-face interview with the participant; this questionnaire has been validated before in an older population at high cardiovascular risk from Spain [17]. Energy and nutrient intake were estimated using Spanish food composition tables [18, 19]. Information on self-reported nut intake was derived from the food frequency questionnaire. The questionnaire includes one item regarding the consumption of almonds, peanuts, hazelnuts, pistachios and pine nuts (macadamias, cashews and Brazil nuts are rarely consumed in Spain), and another question specifically inquired about the consumption of walnuts. The dietitians asked the participants if they consumed this food item never, between 1 to 3 times per month, times per week (1, 2 to 4, 5 to 6; three options) or times a day (1, 2 to 3, 4 to 6, >6; four options). For the purpose of the present study, 28 g of nuts was considered to be one serving. Peanuts, almonds, hazelnuts, walnuts, pine nuts, pistachios, Brazil nuts, macadamia and cashews were all considered nuts. In addition, dietitians administered a validated 14-item MedDiet screener designed to assess the degree of adherence to the traditional MedDiet [20]. We used the score of this brief screener to control for the overall dietary pattern, because a higher adherence to the MedDiet among frequent consumers of nuts could introduce confounding. For this purpose, the question about nut consumption was omitted from the brief screener; therefore, a 13-point score was used as a covariate (minimum 0, maximum 13).

Information on mortality was updated once a year by the End-point Adjudication Committee, whose members were blinded to treatment allocation. Different sources of information were used: (1) yearly questionnaires and examinations to all participants, (2) family physicians, (3) yearly review of medical records, and (4) linkage to the National Death Index. Medical records of deceased participants were requested, and the End-point Adjudication Committee adjudicated the cause of the death.

At baseline, questionnaires about lifestyle variables, educational achievement, history of illnesses, and medication use were administered. Physical activity was assessed using the validated Spanish version of the Minnesota Leisure-Time Physical Activity questionnaire [21]. Participants were considered to be diabetic, hypercholesterolemic or hypertensive if they had previously been diagnosed as such, and/or they were being treated with antidiabetic, cholesterol-lowering, or antihypertensive agents, respectively. Trained personnel took the anthropometric and blood pressure measurements. Weight and height were measured with light clothing and no shoes with calibrated scales and a wall-mounted stadiometer, respectively; waist circumference was measured midway between the lowest rib and the iliac crest using an anthropometric tape; blood pressure was measured using a validated oscillometer (Omron HEM705CP; Hoofddorp, The Netherlands) in triplicate with a 5-minute interval between each measurement, and the mean of these values was recorded.

Follow-up time was calculated as the difference between the date of either death or end of follow-up (the date of the last visit or the last recorded clinical event of participants still alive) and the date of recruitment. Extremes of total energy intake (>4,000 or <800 kcal per day in men and >3,500 or <500 kcal per day in women) were excluded from the analysis [22]. Three categories of frequency of nut consumption were considered (never or almost never, 1 to 3 servings per week and >3 servings per week). We used analysis of variance (ANOVA) or the Pearson χ² tests to compare the quantitative or categorical baseline characteristics of the study participants, respectively, across servings of nut consumption. Results were expressed as means ± SD or percentages. Because no interaction was observed between sex and the main outcome, analyses were conducted for men and women together.

To assess the risk of total mortality by frequency of nut consumption, multivariate relative risks were computed using Cox proportional hazard models, and potential confounders were controlled for. All analyses were stratified by the recruitment center. Results are expressed as hazard ratios (HRs) with 95% confidence intervals (CIs). Given the different nutritional composition of walnuts and other nuts [1], we performed separate analyses for the frequency of total nut consumption, walnut consumption, and consumption of nuts excluding walnuts. After the unadjusted model, another model was adjusted for age (continuous), sex and intervention group. Then, a second model, was additionally adjusted for BMI (continuous), current smoking status (never, former, or current smoker), educational level (illiterate/primary education, secondary education, academic/graduate), physical activity (MET-min/day), total energy intake (kcal/day), history of diabetes (yes/no), history of hypercholesterolemia (yes/no), use of oral antidiabetic medication (yes/no), antihypertensive drugs (yes/no), and statins (yes/no). Finally, a third, fully-adjusted model, was additionally adjusted for alcohol intake (continuous, adding a quadratic term), quintiles of consumption of dietary food groups (vegetables, fruits, red meat, eggs, and fish), and adherence to the MedDiet (13-point score). The same models were used to assess the risk of cardiovascular mortality or cancer mortality, also using Cox proportional hazard models. Linear trend tests were assessed assigning the median value to each category of nut consumption and using it as a continuous variable in the various models. We evaluated the interaction between baseline nut consumption (three categories, two dummy variables) and the intervention group (three groups, two dummy variables) by introducing an interaction term with four degrees of freedom in the model. We used Cox regression models to assess the risk of total mortality, cardiovascular mortality and cancer mortality according to the joint categories of total nut consumption and intervention group. Linear trends were also tested. We had yearly updated information on nut consumption, so to take advantage of this updated information we repeated the analysis using generalized estimating equations to assess the association between repeated measurements of nut consumption and mortality. For each 1-year period we used as exposure the average nut consumption of all repeated measurements from baseline to the beginning of that yearly period.