Adherence to the Mediterranean diet (MD), high in fruits, vegetables and monounsaturated fats, has been associated with lower body mass index. Associations with measured body fat, including regional adiposity, have not been previously investigated. We examined the associations between the alternate Mediterranean diet score (aMED), anthropometry and measured adiposity by dual-energy x-ray absorptiometry (DXA).
This study included 248 healthy females, aged 18–44 years from the BioCycle Study. Each woman’s aMED (range 0–9) was calculated from up to eight 24-h dietary recalls over 1–2 menstrual cycles (>97% had 7 recalls). Multiple linear regression was used to determine whether aMED and its specific components were associated with total and regional adiposity after adjusting for age, race, education, physical activity and energy intake.
Participants had an average (s.d.) aMED of 4.2 (1.7) and percent body fat of 29.5 (6.0)%. Significant inverse associations were found between aMED and all the examined adiposity measures except waist-to-hip ratio. Among the DXA measures, a 1-unit increment in aMED was associated with a 0.06 (95% confidence interval (CI): −0.09, −0.02) lower trunk-to-leg fat ratio (T/L), a measure of upper to lower body fat. In an analysis examining T/L as an outcome with the separate components of the aMED, T/L was lower with increased legume consumption (β=−0.280, 95% CI: −0.550, −0.010) but was higher with increased consumption of red and processed meat (β=0.060, 95% CI: 0.002, 0.117).
Adherence to the aMED was associated with lower total and regional adiposity, adding to the mounting evidence of the health benefits of the MD.
A dramatic increase in obesity in the United States has occurred over the past 20 years.1 In 2009–2010, 32% (95% confidence interval (CI): 29–36%) of reproductive aged US women (20–39 years) were obese (body mass index (BMI) 30 kg/m2).2 Obesity among women of reproductive age carries its toll with the increased risks of infertility, pregnancy complications as well as adverse offspring outcomes.3, 4
Dietary intake has a pivotal role in the rising obesity rates. In particular, examination of dietary patterns, rather than individual nutrients or foods, can capture interactions between separate dietary components.5 One dietary pattern that has been explored in studying obesity is the Mediterranean diet (MD). The traditional MD reflects food patterns found in Greece and Southern Italy in the early 1960s, and is characterized by an abundant intake of plant foods (vegetables, legumes, fruits, nuts, cereals) with olive oil as the main source of fat, a high-to-moderate intake of fish, a low-to-moderate consumption of eggs, dairy products (mainly cheese and yogurt) and poultry, a low intake of red meat and a regular but moderate intake of alcohol (mainly wine during meals).6, 7
Previous studies examining the association between the MD and obesity have mainly assessed BMI and/or waist circumference, as surrogate measures of total and regional adiposity.8 Associations between the MD pattern and measured body fat, including regional adiposity, however, have not been previously investigated among women of reproductive age. Regional adiposity, independent of overall obesity, is a well-recognized risk factor for cardiovascular disease, type 2 diabetes, gestational diabetes and mortality.9, 10 We thus examined the association between the MD, total and regional adiposity by dual-energy X-ray absorptiometry (DXA) and replicated previous findings of the associations between the MD and anthropometry measures using data from healthy premenopausal women participating in the BioCycle Study.
Materials and methods
The BioCycle Study was originally designed to study the association between endogenous sex hormones and oxidative stress.11 Between 2005 and 2006, 259 healthy premenopausal women 18–44 years of age from western New York were followed for 1 (n=9) or 2 (n=250) menstrual cycles and attended up to eight clinic visits per cycle (93.4% with 7–8 visits in cycle 1, 95.6% with 7–8 visits in cycle 2). The visits corresponded to specific menstrual cycle time points with the most hormonal variation, approximately days 2, 7, 12, 13, 14, 18, 22 and 27 of a 28-day cycle adjusted for cycle duration. Cycle visits were routinely scheduled in the morning and the study clinic was open 7 days per week. Main sources of recruitment included study flyers and friends or family referrals who knew about the study or were current study participants.
Participants were not eligible to participate if they were currently using oral contraceptives, vitamin and mineral supplements, or prescription medications; were pregnant or breastfeeding in the past 6 months; had a diagnosis of a chronic medical condition such as metabolic disorders or gastrointestinal diseases; reported a BMI of <18 or >35 kg/m2 at screening; or were on current dietary restrictions for weight loss or for medical reasons. Detailed study methods have been published elsewhere.11 Only participants who had body composition measurements by DXA during their final study visit are included in the current analyses (n=248; 96%). The University at Buffalo Health Sciences Institutional Review Board (IRB) approved the study and served as the IRB designated by the National Institutes of Health for this study under a reliance agreement. All participants provided written informed consent.
Trained study personnel performed anthropometric measures at the baseline visit according to standardized protocols. Height was measured by a fixed Stadiometer while weight was measured on a calibrated balance scale. BMI was calculated by dividing weight in kilograms by height in meters squared (kg/m2). Waist, hip and thigh circumferences were measured in centimeters to the nearest mm with an inelastic measuring tape. Waist circumference was measured at the end of normal expiration midpoint between the iliac crest and the lowest lateral portion of the rib cage. Hip girth was measured at the level of the symphysis pubis and the maximal protrusion of the gluteal muscles. Thigh circumference was measured at the midpoint between the inguinal crease and the proximal border of the patella. Skinfold thicknesses were measured at the triceps (midpoint between the acromion process of the scapula and the inferior margin of the olecranon process of the ulna), thighs (midpoint between the inguinal crease and the proximal border of the patella), subscapular (below the angle of the scapula) and suprailiac (midaxillary line above the iliac crest) to the nearest 1 mm using a Harpenden skinfold caliper. All the anthropometric measurements were done twice and the mean values were obtained from the two measurements.
Certified radiology technicians performed body composition assessments at the final clinic visit (n=248) using DXA (Hologic Discovery Elite, Hologic, Waltham, MA, USA). Ongoing QC and daily phantoms were used to monitor any machine drift. Whole-body assessment (distribution of bone, fat and lean mass) was performed. Total and regional (arms, trunk and legs) bone mass, percent fat and regional fat mass were assessed. Trunk and leg fat were used to compute trunk-to-leg fat ratio (T/L), as a measure of upper to lower body fat. Percent leg fat was calculated as the average amount of fat in both legs divided by total body fat × 100.
Twenty-four hours recall
Twenty-four hours dietary recalls were administered on approximately days 2, 7, 14 and 22 of the menstrual cycle. As such, women could have up to eight 24-h recalls if they completed two cycles. The 24-h recalls were conducted by trained and certified staff through a computerized interview at the time of the morning clinic visit or after 8 pm the prior evening (after fasting has begun) and were administered either by telephone or in person. The Nutrition Data System for Research software, version 2005 (Nutrition Coordinating Center of the University of Minnesota, Minneapolis, MN, USA) was used to derive nutritional information. A validated two-dimensional visual aid helped participants estimate accurate information regarding portion sizes.12 In all, 97.2% of the participants with DXA measures completed more than six 24-h dietary recalls during the two cycles (7 recalls: 21.4%; 8 recalls: 75.8%). Energy intake was averaged across all available recalls per participant.
Alternate Mediterranean diet score
The alternate Mediterranean diet score (aMED),13 a scale adapted from the traditional MD score developed by Trichopoulou et al.,14 is based on the dietary intake of nine components including: vegetables (excluding potatoes), legumes, fruit, nuts, whole grains, red and processed meat, fish, alcohol and the ratio of monounsaturated fat to saturated fat. The aMED ranges between 0 and 9 with higher scores implying greater adherence. Each dietary component receives a score of 1 if consumption exceeds the median intake except for: the red and processed meat component (a point is scored if consumption is less than the median intake) and the alcohol component (a point is scored if intake is between 5 and 15 grams per day). The median intake is derived from the distribution of the dietary intake components of the cohort under study.14
As each participant had more than one dietary recall (4–8 recalls), we calculated the average of each of the aMED dietary components across all available recalls per participant and subsequently calculated one aMED per subject.
Participants reported demographic information including education, race and smoking by questionnaire. The International Physical Activity Questionnaire was administered at baseline and was used to classify physical activity into high, moderate or low.15
Descriptive analyses are presented as means±s.d. for normally distributed continuous variables, and as absolute numbers and frequencies for categorical variables by aMED category. χ2-Test or Fisher’s exact test were used to examine the associations between categorical variables. Differences of mean aMED components and of mean adiposity measurements across the aMED categories (0–3, 4–5, 6–9) were evaluated with analysis of variance. Analysis of variance was also used to examine pairwise contrasts in mean adiposity measurements across the aMED categories. We then used multiple linear regression models to test the associations between the aMED and the different adiposity outcomes (percent body fat, percent trunk fat, percent leg fat, T/L, sum of skinfolds, waist circumference, hip circumference, waist-to-hip ratio (WHR), BMI) adjusting first for average energy intake across the dietary recalls (kcal) and in subsequent models additionally adjusting for age (continuous), physical activity (high, moderate, low), education (high school or less/GED, some college, associates or bachelor degree, graduate degree or higher) and race (white, black, other). The choice of covariates was based a priori on what was previously reported in the literature. Owing to the increased risk of morbidities associated with regional obesity, we further examined the average intake of the individual aMED components by tertiles of T/L and evaluated the associations between each aMED component (independent) and T/L (dependent, continuous) using separate multiple linear regression models adjusting for average energy intake, age, physical activity, education and race. The homogeneity of variance and the independence of the errors assumptions for performing these tests were met. Additionally, our results are robust to the normality of errors assumption because of our large sample size.16 Results from the linear regression models are presented as β-coefficients and 95% CI. All reported P-values are based on a two-sided test with a P-value <0.05 considered statistically significant. SAS version 9.2 (SAS Institute Inc., Cary, NC, USA) was used for all analyses.
Overall, the mean age of this cohort was 27.5 years, 35% had a low aMED (score 0–3) while 22% had a high score (score 6–9). A larger proportion of white women reported a high aMED than women of other race (P=0.07) (Table 1). More than 50% of the participants reported high levels of physical activity with a larger proportion of these women reporting a low aMED (P=0.04). Energy intake increased with increasing aMED. Of the macronutrients, a higher aMED was associated with slightly greater carbohydrate intake but no difference in protein intake. The type of fat consumed significantly differed, with lower saturated and higher polyunsaturated fat among those with a high aMED. As expected, the individual aMED components were directly associated with the aMED in the hypothesized directions.
Women with moderate and high aMED had a significantly lower BMI, smaller waist and thigh circumferences measured at baseline than women with low aMED (Table 2). All of the examined DXA measures except for percent leg fat showed an inverse association with the aMED (P<0.05). Hip circumference and leg fat mass were only significantly different when comparing the moderate aMED to the low aMED group (Table 2). Associations remained significant after adjusting for age, race, education, physical activity and total energy intake (Table 3). An increment of 1-unit in the aMED was associated with a 0.67% (95% CI: −1.11, −0.24) lower percent body fat, a 0.85% (95% CI: −1.39, −0.32) lower percent trunk fat and a 0.06% (95% CI: −0.09, −0.02) lower T/L ratio.
Intake of red and processed meat increased with the increase in tertiles of T/L (mean intake (s.d.)=1.3 servings/day (1.1) in the lowest tertile, 1.5 servings/day (1.2) in the moderate tertile, 1.8 servings/day (1.2) in the highest tertile of T/L). Significant differences were also noted for % of saturated fatty acid (%SFA) with the highest tertile of T/L having the highest %SFA intake (Table 4). When the associations between the individual dietary components of the aMED and regional adiposity as a measure of T/L were evaluated in linear regression models adjusting for age, race, education, physical activity and total energy intake, a high consumption of red and processed meat products and low intake of legumes were significantly associated with increased regional adiposity (P=0.04) (red and processed meat β=0.060, 95% CI: 0.002, 0.117; legumes β=−0.280, 95% CI: −0.550, −0.010) (Table 5). Mutually adjusting for the other aMED dietary components reduced the associations with T/L (red and processed meat β=0.052, 95% CI: −0.011, 0.114; legumes β=−0.25, 95% CI: −0.57, 0.08; P=0.1 data not shown).
As energy intake was significantly related to the aMED and given that the MD might be inherently positively associated with energy intake,17, 18, 19 we also conducted analyses without controlling for energy intake. No qualitative changes were noted in the results (data not shown).
In this cross-sectional analysis, following the principles of a Mediterranean dietary pattern was significantly associated with lower values of total and regional adiposity by DXA and by anthropometry. This association held even after adjusting for potential confounders including age, education, race, physical activity and energy intake.
To our knowledge our findings represent the first demonstration examining the relationship between adherence to the aMED and a direct assessment of total and regional body fat distribution as measured by DXA. These findings are in agreement with previous studies, despite methodological differences (types of study designs, adiposity measurement methods, study populations), showing that following the aMED could potentially be associated with a favorable body weight. Nine cross-sectional studies, of which six examined BMI and three examined both BMI and waist-to-hip ratio/waist circumference, have shown inverse associations with the above adiposity outcomes among study participants with better adherence to the MD.13, 18, 19, 20, 21, 22, 23, 24, 25 Similarly, three of five cohort studies, with follow-up ranging between 2.4–9 years, showed that subjects with better adherence to the MD were less likely to gain weight or to develop overweight/obesity than were individuals with low adherence while the other two cohort studies did not demonstrate any significant weight change.26, 27, 28, 29, 30
The potential physiological mechanisms explaining the protective role of the MD on obesity, the metabolic syndrome and its components as well as cardiovascular disease have been reviewed elsewhere. Briefly, the overall anti-inflammatory and anti-oxidant impact of the MD and the effects of its individual components including olive oil, fruits, vegetables, whole grains and fish offer potential explanations for its protective effects.31, 32, 33 Particularly essential in the prevention of weight gain is the high consumption of dietary fiber with its satiation impact, coupled with a low degree of energy density of the MD overall.33 In a meta-analysis of over 500 000 subjects, a decrease of 0.42 cm (95% CI: −0.82, −0.02) in waist circumference has been reported among subjects with high adherence to the MD.33, 34 This is particularly important for the population under study. Women of reproductive age owing to the childbearing process, are at high risk for obesity development and excessive weight gain during pregnancy with preferential centralized distribution of adipose tissue, a risk factor for a wide range of chronic diseases.35, 36, 37 Our study results support a strong inverse association between measured truncal adiposity and adherence to the aMED implying that not following such a pattern of dietary consumption may be potentially associated with a preference in fat distribution. In a randomized crossover study of overweight or obese men (aged 24–49 years, BMI 25.5–31.3 kg/m2), a significant loss of weight and fat mass resulted after 4 weeks when SFA-rich diet was substituted with predominantly monounsaturated fatty acid-rich diet. Men on the SFA-rich diet accumulated fat mass predominantly on the trunk instead of the limbs.38 The aMED heavily reflects on the fatty acid component of this diet through mainly four categories including meat, nuts, fish and monounsaturated fatty acid to SFA ratio. Interestingly, a food group analysis with T/L shows a difference in consumption for red and processed meat and consumption of legumes. High red and processed meat consumption have been previously associated with increased cardiometabolic risk independent of other dietary factors.39, 40 The lack of an appreciable association with the majority of the individual dietary components of the aMED, might be attributed to the inability of the individual components to capture the synergistic or interactive cumulative effects on adiposity detected through the diet score.14
The findings of this study should be interpreted in light of the study’s strengths and limitations. As this study had a cross-sectional design, we are restricted in terms of making an inference on the time sequence of all associations. The BioCycle study, however, excluded women on a diet for weight loss or for other medical reasons rendering the likelihood of reverse causality less likely to occur. We also had strict inclusion criteria restricting the generalizability of our findings to all US women. Although we used non-objective measurements in assessing physical activity, the instrument we used has previously demonstrated acceptable reliability and validity.15 Strengths of the current study include the use of the 24 h dietary recalls with repeated nutritional assessments (>97% had 7 recalls) over a period of 2 months rendering misreporting of consumed food items unlikely.
In conclusion, we confirmed previous findings that showed an inverse relation between the aMED and anthropometric indices, including waist circumference, waist-to-hip ratio and BMI, and expanded those findings to show the association with total and regional body fat distribution as measured by DXA. These results in addition to the mounting evidence of the health benefits of the Mediterranean diet raise awareness to promote such a lifestyle in dietary habits to combat the worldwide obesity epidemic.
This work was supported by the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health (Contract no. HHSN275200403394C). The authors thank the BioCycle working group for comments and feedback on this work.
About this article
Critical Reviews in Food Science and Nutrition (2018)