INTRODUCTION
Hypertension is a known risk factor for cardiovascular diseases. Therefore, its high prevalence worldwide is a concern. Prevention is reached through lifestyle modifications 1) (2. Hence, decreasing sodium chloride and cholesterol intake and increasing fruit and vegetable consumption are recommended by the European Society of Cardiology 3, alongside with weight reduction, regular physical activity, restricted alcohol consumption and smoking cessation. However, the intake of nutrients and specific food groups is only one aspect of dietary behavior. Another aspect are meal intake behaviors (MIBs) like eating frequency (EF), irregular eating, snacking or skipping meals. These are related to specific lifestyles and dietary patterns and, therefore, might contribute to the development of cardiovascular risk factors 4) (5) (6) (7) (8. However, only a few studies have explored the associations between MIB and blood pressure (BP) 9) (10) (11) (12) (13) (14) (15) (16) (17.
In the case of EF, a systematic review of weight-loss and maintenance intervention studies carried out by Palmer et al. 15 suggested no associations between EF and BP. Moreover, different cross-sectional studies did not find any associations 16) (17 with EF. By contrast, Edelstein et al. 9 showed that systolic blood pressure (SBP) was lower in those subjects who ate more frequently. This was confirmed by Kim et al. 12, who observed an inverse association between EF and BP. Furthermore, a longitudinal study 10 evaluating 115 non-diabetic men and women found that a high eating frequency was associated with decreased systolic (SBP) and diastolic BP (DBP).
Altogether, the results are inconclusive, suggesting either no associations or negative associations between EF and SBP/DBP. This might be due to different methodological limitations: a) EF does not differentiate between meals and snacks; and b) the definition of meals, snacks or even eating occasions is inconsistent across the literature 18) (19. For example, most studies consider only breakfast, lunch and dinner to be meals, coding the rest of the eating occasions as snacks 4) (6. However, most traditional meal patterns consist of more than three meals per day 18) (20. In Spain, for example, in addition to the three main meals, two smaller eating occasions occur, and they are perceived as meals rather than snacks 20. They occur between breakfast and lunch (a forenoon meal, "media mañana") and between lunch and dinner (an afternoon meal, "merienda").
Based on the above mentioned, the present study aims to determine the associations of several MIBs, such as the intake of the three main meals, a forenoon meal, an afternoon meal and snacking between regular meals, with SBP and DBP in a sample of Spanish adults.
MATERIALS AND METHODS
PARTICIPANTS
Data from 1,314 adults (63.2% women and 36.8% men) aged 20-79 years (mean age 57.8 ± 14.9 years) were examined in a cross-sectional study. The survey was carried out in 2008 in four Spanish cities (Madrid, Las Palmas, Seville and Valencia) during a cardiovascular health event organized by the Fundación Española del Corazón and the Sociedad Española de Cardiología. A random sampling of participants was not carried out since the main purpose of these events was to promote the prevention of cardiovascular diseases and to screen for cardiovascular risk factors. Hence, all the volunteers were accepted. Participants were included in the study after completing a guided questionnaire, as well as measurements of waist circumference (WC) and BP. Technicians used standardized anthropometric instruments during measurements, following the recommendations of the International Biological Program 21. The study was approved by the Ethics Committee of the Fundación Española del Corazón, and conducted according to the guidelines in the Declaration of Helsinki 22. Signed consent forms were obtained from all participants.
MEASUREMENTS
A digital tensiometer (Visomat(r) Comfort 20/40) was used to assess SBP (mmHg) and DBP (mmHg). Participants were measured while sitting and after having completed the questionnaire, the intention of which was to achieve a calm state. These measurements were repeated when values were beyond the normal range: from < 90/60 mmHg or > 140/90 mmHg for SBP/DBP 23. WC (cm) was measured midway between the last rib and the upper edge of the iliac crest, using a non-stretchable tape.
MEAL INTAKE BEHAVIORS
The four MIBs investigated (the intake of all three main meals, having a forenoon meal, having an afternoon meal and snacking) were assessed by means of short self-reported questions. Participants were asked about the meals consumed during the day, given the following meals to choose from: breakfast, forenoon meal, lunch, afternoon meal and dinner (1 = yes/0 = no). The intake of all three main meals was confirmed when the three questions on breakfast, lunch and dinner were answered positively. The habit of snacking between their regular meals was estimated by the question "Do you snack between meals?" (1 = yes/0 = no), immediately after asking about their meal intake.
CONFOUNDERS
Known risk factors for hypertension, such as sex, age, individual risk factors and dietary intake, were considered as potential confounders of the MIB-BP association. Individual risk factors were assessed using self-report questions. Participants were asked whether they currently drank alcohol (1 = yes/0 = no) or smoked (1 = yes/0 = no). Additionally, they had to indicate their level of physical activity during spare time. They could select from the following options: sedentary (reading, watching television), light exercise (minimum effort: yoga, walking, fishing), moderate exercise (minimum four hours a week: hiking, cycling, gardening) and intensive exercise (high heart rate: running, football, swimming). For the analysis, moderate and intensive exercises were grouped into one category.
Furthermore, the data from a short food habits questionnaire was used to create two scores that reflect dietary intake. First, the Achievement of the Dietary Guideline Score (ADGS) was based on five short questions about the usual daily intake of: a) meat, fish and eggs (responsive options: zero to five or more servings); b) milk and dairy products (zero to three servings); c) fruit (zero to five or more servings); d) vegetables (zero to three servings); and e) bread, pasta, rice and cereals (zero to five or more servings). One point was obtained for each food group when the serving recommendations of the Spanish dietary guidelines, as presented by Salvador et al. 24, were reached: a) meat, fish and eggs (less than three servings); b) milk and dairy products (more than one serving); c) fruit (more than one serving); d) vegetables (more than one serving); and e) bread, pasta, rice and cereals (more than three servings). By adding up the points obtained, the score ranges from 0 to 5 points, and it is categorized into 1 = very low (0-1 points), 2 = low (2 points), 3 = middle (3 points), and 4 = high (4-5 points). The second score (Unhealthy Habit Score [UHS]) was based on the regular consumption of: a) fatty foods; b) ready-made meals; c) salty foods; d) adding salt to already prepared meals; and e) intake of sugary drinks during a meal. It was assessed by the results of a short question about whether or not each habit was performed regularly (1 = yes/0 = no). The UHS was determined by totaling the points obtained for each question answered with yes (the range is 0 to 5 points), and then recoding them into 1 = very low (0 points), 2 = low (1 point), 3 = medium (2 points) and 4 = high (3-5 points).
STATISTICAL ANALYSIS
Continuous variables were tested for normality and described using means and standard deviation, whereas categorical variables were represented by frequencies. First, the association between the confounders and the dependent variables (SBP and DBP) was assessed by means of linear regression, and adjusted for the investigated MIB. Second, multiple linear regression models were used to examine the associations between the MIBs (the intake of the three main meals, having a forenoon meal, having an afternoon meal and snacking) and SBP/DBP. Model 1 described the analysis adjusted by sex, age, smoking, alcohol consumption, and physical activity during leisure time. Additionally, this model included the MIBs not considered as the main independent variable in each regression, controlling for the mutual effect between them. To examine whether the associations were mediated by dietary intake, represented by ADGS and UHS, or WC, they were introduced in the second (Model 2) and third (Model 3) model, respectively. Finally, Model 4 included all the confounders in the regression. Statistical analysis was conducted using the software package SPSS 17, considering p-value < 0.05 as statistically significant.
RESULTS
Table I presents the sample characteristics according to the investigated MIBs: the intake of all three main meals, having a forenoon meal, having an afternoon meal and snacking. The majority of the sample consumed all three main meals. More than one third of the participants took a forenoon meal and nearly half of them an afternoon meal, whereas only a fourth of the individuals snacked between their regular meals. SBP was significantly lower among the participants who usually had a forenoon meal, an afternoon meal and those who snacked between the regular meals. In contrast, subjects who had all three main meals showed a significantly higher SBP. Additionally, those who had a forenoon meal and an afternoon meal also presented lower DBP.
n: Number; SD: Standard derivation; DBP: Diastolic blood pressure; SBP: Systolic blood pressure; WC: Waist circumference.
Table II shows the associations between SBP/DBP and the confounders (sex, age, smoking, drinking alcohol, physical activity during spare time, the ADGS and UHS). Subjects with an increased WC had significantly higher BP. SBP was lower among men and individuals with moderate to intense physical activity, whereas SBP increased with age. In contrast, DBP did not show any associations with the confounders.
DBP: Diastolic blood pressure; SBP: Systolic blood pressure; WC: Waist circumference; aReference categories: Sex (men), smoking (no), alcohol consumption (no), physical activity during spare time (sedentary), Achievement of the Dietary Guideline Score (very low), Unhealthy Diet Score (very low). bSignificant association p ≤ 0.05.
Finally, table III shows the associations between the MIBs and SBP/DBP. After adjusting for the confounders, MIB not being considered to be the main independent variable, having an afternoon meal was directly associated with SBP and DBP. The participants who had an afternoon meal had significantly lower SBP and DBP. Those associations were attenuated when the ADGS, UHS and WC were included in the regression models, yet, the association remained significant. In contrast, making all three main meals, the intake of a forenoon meal and snacking did not show associations with SBP and DBP.
aModel 1: Adjusted for sex, age, individual risk factors (smoking, alcohol consumption, physical activity in leisure time); MIB not considered to be the main independent variable. bModel 2: Model 1 and Achievement of the Dietary Guideline Score, Unhealthy Habit Score. cModel 3: Model 1 and WC. dModel 4: Model 1 and Achievement of the Dietary Guideline Score, Unhealthy Habit Score and WC. DBP: Diastolic blood pressure; SBP: Systolic blood pressure; WC: Waist circumference. eReference category: no. fSignificant association p ≤ 0.05.
DISCUSSION
In this study, we used linear regression models to assess associations between the MIBs (intake of the three main meals, forenoon meal, afternoon meal and snacking between the regular meals) and SBP/DBP. Our findings showed that subjects who consumed an afternoon meal seemed to have lower BP independently of other variables such as sex, age and individual risk factors. However, the association was attenuated through dietary intake, represented by ADGS and UHS, and WC. On the other hand, the rest of the investigated MIBs were not related to SBP/DBP.
To the best of our knowledge, the negative direct association between the afternoon meal and SBP/DBP is a unique result, as meals in addition to breakfast, lunch and dinner are underexplored for several reasons. First, in previous studies extra meals or small meals are usually examined as snacks, and therefore are considered the same as snacking in general 4) (6. For example, Kim et al. 12 found snack frequency, but not meal frequency, to be inversely associated with BP. However, the used meal definition included only breakfast, lunch and dinner, whereas other eating occasions were considered as snacks. Second, an afternoon and a forenoon meal are meals traditionally taken in Spain.
However, they are much less common in other populations. For instance, Barba et al. 25 studied the association between EF and BP in children from southern Italy. When defining EF, they included, in addition to the three main meals, a forenoon meal, an afternoon meal and milk before bed.
The possible mechanisms by which the intake of an afternoon meal might reduce SBP and DBP are unclear, especially, as there was no observed association with the forenoon meal. However, we believe that the afternoon meal might somehow affect the intake of the subsequent meal. In the case of the forenoon meal this is lunch, whereas for the afternoon meal it is dinner. The amount of energy consumed during dinner might be one of the highest of the day, as observed by several studies from different countries. Additionally, previous investigations showed a decline in insulin sensitivity throughout the day, reaching its lowest point in the evening 27. Hence, the intake of an afternoon meal may result in a decreased energy intake during dinner and thus, perhaps, it improves insulin metabolism. Yet, given the observational nature of the study, we were not able to test this assumption. However, an intervention study carried out by Chapelot et al. 18 showed a lower dinner intake in those participants who usually had a traditional French mid-afternoon eating occasion. This occurred even though the participants who never make such a small meal in the afternoon ate a snack out of the same food items as provided for this afternoon meal. However, after a five-year follow-up, Karatzi et al. 10 found a direct link between high EF and a lower rate of SBP and DBP, independently of the homeostatic model assessment-insulin resistance (HOMA-IR). Nevertheless, they suggest that over the long term a reduced insulin concentration might benefit cardiovascular health. As the afternoon meal is part of a traditional Spanish meal pattern, the long-term effect postulated might have acted over metabolism. Additionally, it is important to mention that in the study by Karatzi et al. 8 participants were screened to have no conditions, such as diabetes mellitus, liver or endocrine diseases, that may impact blood pressure. In our study, participants were enrolled regardless of the diseases they had.
In addition, with the use of regression models we could assess the role of dietary intake, represented by ADGS and UHS, as well as WC on the associations between MIBs and SBP/DBP. In the first place, we observed that the association between the intake of the afternoon meal and BP was unconnected to these two confounders. This is in line with the study carried out by Karatzi et al. 10, who also found that the association of EF with BP after a five-year follow-up remained significant after the adjustment of BMI and energy intake. In contrast, Barba et al. 25 found an association between EF and BP in children from southern Italy that was not independent from BMI, which is also confirmed by Kim et al. 12 in an adult population. Secondly, although the association was still significant, attenuation could be observed when WC and dietary intake were included in the regression models, which may also indicate an indirect pathway of the MIB-BP association. Both confounders are known risk factors of high BP, although the metabolic pathways underlying the mechanism are still under discussion. The mechanisms are possibly multifactorial, such as endothelial dysfunction, alteration of the nervous system and kidney function, and the modification of the balance of specific hormones like insulin and leptin 28) (29.
This study had several limitations. First, the sample was neither randomly selected nor representative of the Spanish population. The data were collected during events whose objectives were targeting cardiovascular risk factors among the participants and providing information about current lifestyle. Therefore, the study was biased towards certain risk groups that might be more sensitive to their own health, as observed by the higher participation of women and older people. Second, causal relationships could not be established because of the cross-sectional design of the study, even though the analysis was carefully adjusted. However, residual confounding cannot be ruled out. For example, our results might be biased for not having considered the use of antihypertensive medication. This might have led to underestimation of the MIBs-BP association and the presence of disease with an impact on cardiovascular health, such as diabetes mellitus or kidney diseases. Third, the questionnaire used to assess dietary intake was not previously validated.
In contrast, the study had various strengths. First, the use of linear regression models allowed us to deepen into the nature of the associations. Second, WC was measured directly rather than self- reportedly, thus avoiding response errors and underestimations. Finally, a trained person guided the participants through the questionnaire, guarding against wrong interpretations of questions, as well as memory gaps and inexactness.
In conclusion, we found that the intake of an afternoon meal was directly associated with lower BP. This represents a novel result as MIBs, which had rarely been studied before, were investigated. The results might easily be included in dietary advisories. However, in order to confirm the findings, further population-based studies by means of validated measurements of dietary pattern and confounders are needed.