Change to a healthy diet in people over 70 years old: the PREDIMED experience

Purpose It is difficult to change dietary habits and maintain them in the long run, particularly in elderly people. We aimed to assess whether adherence to the Mediterranean diet (MedDiet) and cardiovascular risk factor were similar in the middle-aged and oldest participants in the PREDIMED study. Methods We analyzed participants belonging to the first and fourth quartiles of age (Q1 and Q4, respectively) to compare between-group differences in adherence to the nutritional intervention and cardiovascular risk factor (CRF) control during a 3-year follow-up. All participants underwent yearly clinical, nutritional, and laboratory assessments during the following. Results A total of 2278 patients were included (1091 and 1187 in Q1 and Q4, respectively). At baseline, mean ages were 59.6 ± 2.1 years in Q1 and 74.2 ± 2.6 years in Q4. In Q4, there were more women, greater prevalence of hypertension and diabetes, and lower obesity and smoking rates than the younger cohort (P ≤ 0.001, all). Adherence to the MedDiet was similar in Q1 and Q4 at baseline (mean 8.7 of 14 points for both) and improved significantly (P < 0.01) and to a similar extent (mean 10.2 and 10.0 points, respectively) during follow-up. Systolic blood pressure, low density–lipoprotein cholesterol, and body weight were similarly reduced at 3 years in Q1 and Q4 participants. Conclusion The youngest and oldest participants showed improved dietary habits and CRFs to a similar extent after 3 years’ intervention. Therefore, it is never too late to improve dietary habits and ameliorate CRF in high-risk individuals, even those of advanced age. Registration The trial is registered in the London-based Current Controlled Trials Registry (ISRCTN number 35739639). Supplementary Information The online version contains supplementary material available at 10.1007/s00394-021-02741-7.


Introduction
Cardiovascular disease (CVD) has the highest incidence and prevalence in elderly persons, in whom it is associated with high morbidity and mortality [1,2]. In Europe, CVD incidence is predicted to increase in the near future because of population ageing [3,4]. In addition, the elderly frequently suffer multimorbidities, and a significant proportion can be classified as frail [5]. Frailty shares some physiopathological mechanisms with CVD and both conditions are tightly interrelated, as frailty can worsen CVD and CVD may precipitate frailty [6,7]. A healthy dietary pattern, such as the Mediterranean diet (MedDiet) is associated with both lower rates of non-communicable diseases, including CVD [8,9] and frailty [10,11].
CVD is a main component of multimorbidity, and may be prevented and managed by pharmacological and nonpharmacologic therapies, such as a healthy diet, increased physical activity, and abstention from smoking [12]. In fact, lifestyle changes are the cornerstone of CVD prevention and are recommended in all international guidelines [12,13]. The main problem with lifestyle measures is that a change in patient behaviour is mandatory to transform unhealthy habits to healthy ones and maintain them in the long run, which is particularly difficult in the elderly, due to several factors such as lower appetite, food changes, declining physical function, cooking for one, shopping for one, food cost, etc. [14][15][16][17].
The PREvención con DIeta MEDiterránea (PREDIMED) randomized trial demonstrated that a MedDiet supplemented with extra-virgin olive oil (EVOO) or mixed nuts can improve cardiovascular risk factor (CRF) control [18,19] and reduce CVD incidence by nearly 30% in older individuals at high cardiovascular risk compared to advice on a low-fat diet (LFD) [9]. However, it is unknown whether adherence to the MedDiet was similar in the youngest and oldest participants in the PREDIMED trial and also whether the beneficial effect on CRFs differed in younger and older individuals, since in a previous PREDIMED study, short and long-term predictors of dietary changes were analyzed, but age was not a clear predictor in either of these two analyses [17].

Study participants and design
The PREDIMED study (www. predi med. es) was a nutritional intervention-based randomized trial, single-blind, multicenter (11 centres throughout Spain) study carried out on 7447 participants at high vascular risk, but no CVD at enrolment, conducted from October 2003 to December 2010. The aim of this primary prevention study was to assess the long-term effects of the MedDiet on incident CVD in men and women aged 55-80 years at high CVD risk. The protocol, methods, design, and eligibility criteria for this study have been reported in detail elsewhere [9]. Briefly, participants were assigned to one of three nutritional interventions by a computer-generated random-number sequence: a traditional MedDiet supplemented with either complementary EVOO or tree nuts and a control diet based on advice to follow an LFD.
In the current study, we analyzed data of participants belonging to the first (n = 1091) and fourth (n = 1187) quartiles of baseline age to compare dietary intake, adherence to the nutritional interventions, and CRF control. We included a total of 2278 participants with complete information on food consumption and nutrient intake, adiposity, and CRFs at baseline and after 3 years of follow-up.

Assessments and intervention
Throughout the study, participants in the MedDiet groups had face-to-face interviews with a dietitian (yearly) and at group sessions (every 3 months) in which they received instructions and written material with information on seasonal Mediterranean foods, shopping lists, weekly meal plans, and cooking recipes for a typical week. They were instructed to follow the allocated diets, with different sessions for each intervention group. During the group sessions with dietitians and according to treatment allocation, participants in the two MedDiet groups were provided with either 1 L per week of EVOO (to consume at least 50 mL/ day) or mixed nuts (30 g/day: 15 g walnuts, 7.5 g hazelnuts, and 7.5 g almonds) at no cost. Participants in the LFD group also had individual and group sessions quarterly with the dietitian, were given information and written material on the LFD (according to the American Heart Association guidelines), and received non-food gifts. None of the three groups received advice on energy restriction or promotion of physical activity.
A validated 14-point MedDiet screener was used to assess adherence to the MedDiet [20], while a nine-item questionnaire was used to evaluate adherence to the LFD 1 3 [9]. Also, at baseline and yearly during follow-up, dietitians administered in face-to-face interviews a 137-item validated food-frequency questionnaire (FFQ) [21], used to assess energy and nutrients using Spanish food-composition tables [22]. The Minnesota Leisure Time Physical Activity Questionnaire and a 47-item questionnaire on education, lifestyle, history of illnesses, and medication use were also administered at baseline and yearly.

Clinical measurements
Trained personnel performed anthropometric measurements. Height and weight of volunteers were measured using a wall-mounted stadiometer and calibrated scales, respectively. Waist circumference was measured midway between the lowest rib and the iliac crest using an anthropometric tape. Blood pressure (BP) was measured in triplicate with a validated semiautomatic oscillometer (Omron HEM-705CP, Hoofddorp, Netherlands).
In addition, fasting blood and spot urine were obtained and plasma, serum, and buffy coats stored at − 80 °C until assay. The analytes determined in frozen samples of serum or plasma as appropriate were glucose by the glucose oxidase method, cholesterol and triglyceride by standard enzymatic procedures, high-density lipoprotein (HDL) cholesterol after precipitation with phosphotungstic acid and magnesium chloride, and low-density lipoprotein (LDL) cholesterol by the Friedewald formula when triglycerides were < 400 mg/dL.

Ethics statement
All participants provided written informed consent to a protocol designed according to the ethical guidelines of the Declaration of Helsinki that had been approved by the institutional review boards of all participating centres. The Institutional Review Board of the Hospital Clinic (Barcelona, Spain), accredited by the US Department of Health and Human Services update for Federalwide Assurance for the Protection of Human Subjects for International (non-US) Institutions (00000738), approved the study protocol on July 16,2002. The trial was registered (ISRCTN35739639).

Statistical analyses
Subjects were stratified into quartiles of age. The first quartile comprised participants ≤ 62 years old (youngest) and the fourth quartile those ≥ 71 years old (oldest). Baseline characteristics of the participants are expressed as means ± SD) or percentages as appropriate. Kolmogorov and Levene tests were used to assess data normality and skewness. One-factor ANOVA with two factors (treatment group and age quartiles 1-4) was used for continuous variables and χ 2 tests for categorical variables. Analysis of the effects of treatment and age quartile (Q1 vs Q4) was performed using ANOVA for analysis of the baseline visit and ANCOVA adjusted for change from baseline at 3 years. For changes in scores on the 14-item questionnaire of Mediterranean diet adherence in extreme quartiles of age, we used χ 2 for comparisons between age groups, diet groups, changes between age groups, changes between diet groups. We also used the McNemar test to compare between baseline and 3-year values. Analyses were performed using SPSS 20.0. Significance level was set at P < 0.05.

Study population
Of the 7447 participants included, we analyzed data of 2278 with complete information on food consumption, nutrient intake, adiposity, and CRFs at baseline and after 3 years of follow-up (Fig. 1). The first (≤ 62 years, mean 59.6 ± 2.1 years) and fourth (≥ 71 years, mean 74.3 ± 2.6 years) age quartiles were composed of 1091 and 1187 subjects. Baseline characteristics of these participants are summarised in Table 1.
In Q4, there were more women, greater prevalence of hypertension and type 2 diabetes mellitus, and lower prevalence of dyslipidemia, overweight/obesity, smoking, and family history of ischemic heart disease than participants in Q1 (P ≤ 0.006, all). With regard to medication, older participants took more angiotensin-converting-enzyme inhibitors, oral hypoglycaemic drugs, aspirin or other antiaggregants, antidepressants, diuretics, vitamins, or supplements than younger subjects (P < 0.05, all). In Q1 and Q4, participants in the control group (LFD) took more antidepressants and diuretics than those in the two MedDiet groups (P < 0.05, both).

Adherence to MedDiet based on the 14-item questionnaire
As expected, both Q1 and Q4 participants allocated to the MedDiet groups significantly improved MedDiet adherence on 13 of 14 score items (Supplementary Table 1). Almost all participants (~ 97%) in the MedDiet groups used OO as their main culinary fat, whereas this percentage was lower (~ 93%) in the control group (P < 0.001, both). At the end of the 3-year-intervention, ≥ 75% of participants in the Med-Diet groups had appropriate intake of red meat, butter, carbonated beverages, chicken, turkey, rabbit, fish, shellfish, and dishes dressed with sofrito, while optimal daily consumption of fruit and vegetables was achieved by only 60% of the participants.
These changes brought patients' diets closer to the Med-Diet pattern. In fact, after 3 years of intervention, > 75% of participants allocated to the two MedDiet groups fulfilled the criteria for eight of the 14 items evaluated in our MedDiet questionnaire. Dietary improvement was lower in the control group than the MedDiet groups, and consisted mainly in decreased consumption of red meat and butter and moderate increases in vegetables, fish, and white meat (P < 0.05, all).

Changes in intake of selected foods and physical activity
As shown in Table 2, at the 3-year follow-up, MedDietadherence scores were greater (about 1.6-2 points) in Q1 and Q4 participants in both MedDiet groups than those in the LFD group (P < 0.001). Adherence to the MedDiet intervention was similar between the youngest and oldest PRED-IMED participants at the end of the intervention.
In addition, the two MedDiet groups showed good adherence to the supplemented foods (EVOO or nuts, P ≤ 0.001 for both), which was slightly higher in younger than older (P ≤ 0.065, both) participants. Both the Q1 and Q4 groups showed increased consumption of fruit and legumes and decreased consumption of meat and meat products and cereals. Consumption of pastries, cakes, sweets, and alcohol had significantly reduced in both groups (Q1 and Q4, P ≤ 0.012, all). In comparison with Q4, participants in Q1 showed significantly increased consumption of vegetables, fish, seafood, tea, and coffee and significantly decreased consumption of milk and dairy products (P < 0.05, all).
However, the MedDiet groups in Q1 and Q4 differed in some specific foods. As such, these groups reported higher consumption of fruit, legumes, fish and seafood than the control group (P ≤ 0.025, all). In addition, the Q1 group disclosed higher consumption of vegetables and tea (P < 0.05, both) than   Table 1 Baseline characteristics of study subjects stratified by extreme quartiles of age Values are mean ± SD or n (%) as appropriate. ANOVA with two factors (group of treatment and quartiles 1-4 of age) was used for continuous variables and     the Q4 group, while the latter group had higher consumption of milk and dairy products (P = 0.032). Finally, only participants in the Q1 group had significantly increased their physical activity during follow-up, which was significantly higher than those in Q4 (P < 0.001). On the other hand, Q4 participants in the control group had significantly decreased physical activity (P < 0.05).

Changes in energy and nutrient intake in young and old cohorts
The Q1 and Q4 groups had significantly decreased consumption of cholesterol and calcium at the end of the study (P ≤ 0.007). Participants in Q1 had increased intake of magnesium in comparison to Q4 (P = 0.051), while participants in Q4 had higher reductions in sodium than participants in Q1 (P = 0.08). On the other hand, consumption of polyunsaturated fatty acids (PUFAs; P < 0.001), α-linolenic acid (P < 0.001), and marine n-3 FAs (P ≤ 0.002) had increased during the study period in both quartiles (Table 3).
Q1 and Q4 subjects in the two MedDiet groups had reduced consumption of protein intake, total carbohydrates, saturated FAs (SFAs), potassium, and sodium (P < 0.05, all) and increased consumption of fibre, monounsaturated FAs (MUFAs), α-linolenic acid, and marine n-3 FAs (P < 0.001, all). In addition, Q1 and Q4 participants on the MedDiet supplemented with nuts had increased consumption of α-linolenic acid and magnesium (P < 0.001, both). On the other hand, for the LFD in both Q1 and Q4, there were significant reductions in energy, total fat, and calcium intake (P < 0.001). Finally, Q1 participants in the two MedDiet groups showed significantly increased PUFAs (P < 0.001).

Changes in classical cardiovascular risk factors
As represented in Fig. 2 Table 2, systolic BP and HDL-cholesterol levels had decreased in both quartiles (P ≤ 0.002, both) at 3 years, although the change was greater in the Q1 group. In addition, modest reductions in body weight and (P < 0.05) occurred in both quartiles, slightly superior in Q4 (P ≤ 0.04). Likewise, a reduction in serum LDL-cholesterol (P = 0.030) was observed in Q1 and Q4 subjects, although this reduction was higher in the Q4 group. Finally, it is noteworthy that diastolic BP had decreased in Q4, contrary to Q1, whose BP had increased (P < 0.001 and P interaction = 0.01). No significant differences were observed in serum triglycerides, fasting blood glucose, or waist circumference.

Discussion
The present results show that adherence to the nutritional intervention implemented in the PREDIMED trial was similar between older (Q4, ≥ 71 years old) and younger (Q1, ≤ 62 years old) participants and was maintained throughout a 3-year follow-up. Both cohorts had increased their scores on the 14-item MedDiet questionnaire by 1.6-2 points at the end of the intervention. As a result, improvement in CRF control was also similar in the two groups. In concordance, subgroup analyses on the primary outcome of the PREDIMED trial also revealed similar CVD-risk reduction with the MedDiet in participants aged < 70 years and ≥ 70 years [9]. While comparisons of dietary intake and MedDiet adherence with younger participants was not possible due to inclusion criteria (aged 55-80 years), some differences has been observed. Although younger and older subjects reached similar overall adherence to the MedDiet (10.1 vs 10.0 of 14 points, respectively), some between-group differences deserve to be mentioned. At the end of the intervention, a lower proportion of Q4 participants had achieved recommended doses of OO, nuts, fruit, legumes, and commercial sweets, while a higher percentage had reached recommended doses of red meat, dietary milk, and alcohol than Q1 participants.
The National Health and Nutrition Examination Survey and other epidemiological studies have revealed that intake of beneficial nutrients, such as complex carbohydrates, fibre, MUFAs, and PUFAs, is reduced in older individuals, while intake of unhealthy nutrients, such as SFAs, is increased [23][24][25]. Data from our study showed that independently of age, allocation to the two MedDiet groups resulted in significant reductions in SFA intake and increased intake of MUFA and fibre, while PUFA intake increased only in participants allocated to the MedDiet supplemented with nuts. Higher fibre intake can be related to healthy dietary changes, such as higher consumption of fruit, vegetables, legumes, and nuts. In fact, higher intake of MUFAs, PUFAs, and fibre can have a protective effect against such age-related disorders as cognitive decline, CVD, diabetes, and cancer, as well as development of frailty [8-11, 18, 26-30]. Furthermore, it should be noted that while both MedDiet groups from the Q4 group significantly increased marine ω 3 FA intake, only 3 years. *P < 0.05 and **P < 0.01 indicates statistical significance by t-test for related samples. BMI body mass index, DBP diastolic blood pressure, HDL high density lipoprotein, LDL low density lipoprotein, SBP systolic blood pressure those allocated to the MedDiet supplemented with nuts significantly increased α-linolenic intake, as expected from the richness of walnuts in this FA. Importantly, intake of longchain PUFAs, such as marine ω 3 FAs and α-linolenic acid, is associated with improved cognition and reduced risk of Alzheimer's disease and other dementias [28,29,31].
Ageing is frequently associated with malnutrition, particularly in frail individuals, due to such factors as hyporexia, decreased saliva production, disturbances in taste and smell, and biological changes, such as alterations in ghrelin and cholecystokinin production, as well as polypharmacy effects [14,16]. Epidemiological studies have shown that compared to young people, the average daily caloric intake is lower in the elderly by approximately 1000 and 700 kcal/ day in men and women, respectively [24]. Also, an observational study reported that a significant proportion of persons aged > 80 years had a daily energy intake < 20 kcal/kg body weight [32]. However, in our study, daily energy intake at the end of the study was lower in Q1 and Q4 participants in the LFD group than the two MedDiet groups, whose intake of protein and fat was maintained or even increased. In other reports on elderly people, animal-protein intake was reduced, which was attributed in part to difficulty chewing and swallowing [33]. It is well established that old people must maintain adequate daily protein intake as a preventive measure to preserve skeletal muscle mass and avoid sarcopenia and frailty [34].
As stated in most guidelines, non-pharmacological measures are the first therapeutic approach to improve control of CRFs and reduce incidence of CVD [5,12,13]. However, there is concern whether elderly persons are capable of improving unhealthy lifestyle habits and maintaining beneficial changes in the long run, probably because changes associated with ageing, such as loss of appetite (reduced taste and smell), loneliness, eating alone, depression, and low income, can influence food choices and dietary habits [14][15][16]. The usefulness of non-pharmacological measures to reduce CVD incidence or improve nutritional status in elderly people has recently been reviewed [35]. While the evidence is of low quality, because it was based on heterogeneous results obtained from small cohorts with short followup periods [35][36][37][38][39][40][41][42], the results are encouraging. Adherence to a MedDiet intervention for 6 months in a cohort of 166 elders (mean age 71 years) was high (85%) and associated with lower BP and improvement in endothelial function [36]. Likewise, adherence to a DASH diet for ≤ 3 months in two cohorts of aged Asian individuals was also high and resulted in lower BP [37][38][39]. Also, a multicomponent nutritional telemonitoring intervention applied for 6 months in elderly people (mean age 78 years) at risk of undernutrition showed good adherence and resulted in improved diet quality and nutritional status [40]. The NU-AGE project conducted on a cohort of 1141 elderly European subjects demonstrated that it is possible to change dietary habits of elders towards a healthier diet that can improve cognitive and bone health [41]. On the other hand, regarding physical activity, preliminary data from the PREDIMED-Plus study have revealed that it is also feasible to increase physical activity in old people in the long run (12 months) [42], which confirms findings from a recent meta-analysis [43]. Other studies, however, have shown negative results [44,45].
Our study has strengths, such as the clinical trial design, repeated data collection, validated FFQ, standardized measurements of clinical and nutritional variables, a relatively large sample (2200 patients), and long-follow-up (3 years). The main limitations are the use of the FFQ may have led to a misclassification of the exposure due to an overestimation of food intake and the fact that dietary data are self-reported. In addition, self-reported questionnaires about diet, physical activity and other medical data can lead to misclassification, which would attenuate the association of the exposure variables with the outcome. Furthermore, potential residual confounding and the lack of generalizability of the results to other populations are limitations in this study. Unmeasured confounders may have distorted results for predictors of dietary adherence, though analyses were adjusted for a wide array of confounders, and a strong confounder unrelated to these characteristics is unlikely. Finally, the findings in our Mediterranean cohort of individuals at high cardiovascular risk cannot easily be extrapolated to other populations.

Conclusion
We report that persons aged > 70 years can improve their dietary habits and adhere in the long term to an enhanced MedDiet in a similar way to younger adult individuals. This goal was reached in part because participants were taught and trained with high intensity by motivated dietitians and received key MedDiet foods for free. As a healthy and high-quality diet, the MedDiet was associated with reduced potency of CRFs to a similar extent in elderly and younger individuals. The benefits of the MedDiet for non-communicable diseases include reduced rates of diabetes and some cancers, lower BP, and improved cognition, as described in other PREDIMED reports [18,19,29,30]. The take-home message is that we should not miss the opportunity to apply such non-pharmacological measures as the MedDiet, which has high efficacy without adverse effects, to improve the overall health of aged people. It is never too late to change dietary habits to achieve healthy ageing.