Epidemiology of metabolic health
Slagter, Sandra Nicole
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2017
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Slagter, S. N. (2017). Epidemiology of metabolic health: Lifestyle determinants and health-related quality of life. Rijksuniversiteit Groningen.
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
5
Hoofdstukpagina-letteromtrek.indd 9 16/11/16 11:05
5
Dietary patterns and physical activity in
the metabolically (un)healthy obese: The
Dutch LifeLines Cohort Study
Sandra N. Slagter Eva Corpeleijn
Melanie M. van der Klauw Anna Sijtsma Linda Swart Corine Perenboom Jeanne de Vries Edith J. Feskens Bruce H.R. Wolffenbuttel Daan Kromhout
Jana V. van Vliet-Ostaptchouk In preparation
5
Hoofdstukpagina-letteromtrek.indd 9 16/11/16 11:05
5
abStract
introduction The diversity in the reported prevalence of metabolically healthy
obe-sity (MHO), suggests that modifiable factors, such as diet and physical activity, may contribute to metabolic health. The aim of this study was to evaluate differences in dietary patterns and physical activity between MHO and metabolically unhealthy obesity (MUO).
methods Data was used of 9,270 obese individuals, aged 30-69 years, of the LifeLines
Cohort Study. Diet, using a 111 item Food Frequency Questionnaire, and physical activity were self-reported. MHO, intermediate obesity and MUO were defined ac-cording to the presence of obesity, number of MetS risk factors and history of cardio-vascular disease. Sex specific associations of dietary patterns, identified by principal component analysis, and physical activity with MHO were assessed by multivariable logistic regression (reference group: MUO), adjusted for among others demographic characteristics, smoking and alcohol use.
results Among 3,442 men and 5,828 women, respectively 10.2% and 24.4% was MHO
and 56.9% and 35.3% MUO. We generated four obesity-specific dietary patterns of which two were related to MHO in women only. In the highest quartile (Q) of the ‘bread, potatoes and sweet snacks’ pattern the adjusted odds ratio (OR) (95% confi-dence interval) for MHO was 0.60 (0.44-0.81). A positive association with MHO was found for the more healthier pattern ‘fruit, vegetables and fish’, with OR 1.31 (1.04-1.64) in Q3 and 1.46 (1.15-1.87) in Q4. In contrast to women, men in the highest tertile of the vigorous physical activity score had a 1.96 (1.45-2.64) OR for MHO. Non-smoking and alcohol use were positively associated with MHO, in both men and women.
conclusion Our results suggest that a healthier diet and vigorous physical activity is
associated with MHO in respectively women and men. Identification of behavioural lifestyle patterns in the obese population may help in pinpointing vulnerable sub-groups and to develop potential strategies improving metabolic health.
5
iNtroductioN
The prevalence of overweight and obesity is increasing in today’s obesogenic environ-ment [1]. Obese individuals are more likely to develop multiple metabolic complications which increases their risk of type 2 diabetes (T2D) and cardiovascular disease (CVD) [2]. However, obesity is a complex and heterogeneous condition with phenotypic variation. Some obese individuals, called the metabolically healthy obese, show no sign of condi-tions associated with the metabolic syndrome (MetS), i.e. impaired glucose metabolism, hypertension and dyslipidaemia [3]. Whether metabolically healthy obesity (MHO) is a truly healthy state remains controversial. Meta-analyses have indicated that adults with MHO have a risk for T2D, CVD [4] and mortality [5] that is intermediate between that of healthy normal weight and unhealthy obese adults. Meaning that even without weight loss, the (cardio)metabolic health can be improved in obese individuals. This is an im-portant finding, given the fact that sustained weight loss is difficult [6, 7].
Interestingly, the metabolically healthy obesity (MHO) phenotype may be modifiable. Several longitudinal studies with up to 10 years of follow-up showed that 43.3-47.6% of those grouped as MHO transitioned to metabolically unhealthy obesity (MUO) [8-10]. Of course, it is partially a transient state due to ageing and the associated adverse meta-bolic changes. However, MHO prevalence does not only differ between age groups, but also between countries. The BioSHaRE-EU Healthy Obesity Project reported that among 28.077 obese individuals from different European countries, the MHO prevalence was lower within higher age groups and lower among men compared to women. Using the same diagnostic criteria, even the age-standardized prevalence of MHO was highly varying between countries, 2-19% among men and 7-28% among women [11]. Thus on top of age, sex and genes, this suggests that other factors are related to the transition from healthy to unhealthy, and possible, in reverse direction, from unhealthy to healthy obesity.
Characterizing the metabolically healthy- and unhealthy obese is of primary impor-tance for medical research and clinical practice, since sizable benefits may still be real-ized by promoting MHO [12]. Which determinants account for the metabolic differences observed between MHO and MUO is uncertain, and particularly data on the role of diet and physical activity are limited. Previous studies on intake of single foods and/or mi-cro- and macronutrients could not find an association with metabolic health subtypes [13-16]. However, nutrients interact with each other in food products and whole diets [17]. Examination of dietary patterns may, therefore, be more suitable to gain insight into the relation between diet and metabolic health.
‘A priori’ dietary scores are based on current knowledge about the role of foods and nutrients in the etiology of disease. They are often developed to assess diet quality based on adherence to nutritional recommendations [18]. For example, in the study of
Cahmi et al. [19], adolescents and women with MHO (19-44 years) had higher scores on the Healthy Eating Index (HEI-2005), which assess diet quality in relation to U.S. National Dietary Guidelines (2005), compared to the MUO individuals. Less is known about exist-ing food consumption patterns within the obese population. An ‘a posteriori’ approach of examining such dietary patterns is factor analysis. With this data-driven technique, dietary patterns are identified based upon intercorrelations between dietary items within the studied population [17]. Although they do not necessarily represent optimal diets for risk assessment, they are an expression of the way how people eat [17] and are expected to be part of a broader pattern of lifestyle factors [18]. Compared to ‘a priori’ dietary scores, dietary patterns derived from factor analysis are more likely to generate new hypotheses, and may improve our insight into possibilities for dietary changes.
The aim of this study was to evaluate differences in dietary patterns and physical activity, between MHO and MUO in the large population-based LifeLines Cohort Study. More specifically, we aimed to 1a) generate obesity-specific dietary patterns, and 1b) ex-amine their associations with demographic- and other lifestyle factors; and 2) compare the dietary patterns and physical activity between MHO and MUO, taking into account among others demographic characteristics, smoking and alcohol use.
mEthodS
lifelines cohort study
LifeLines is a prospective population-based cohort study using a unique three-genera-tion design to study the health and health-related behaviours of 167,729 persons living in the North of The Netherlands. The LifeLines adult population is broadly representative for the adults living in this region [20]. Detailed information on the cohort profile can be found elsewhere [21].
Before study entry, all participants signed an informed consent. The LifeLines Cohort Study is conducted according to the principles of the Declaration of Helsinki and in ac-cordance with the research code of the University Medical Center Groningen (UMCG). The study has been approved by the medical ethics review committee of the UMCG.
For this study we used a subset of the cross-sectional data, collected between 2006 and 2013. Subjects were included in the present study if they had obesity (body mass index (BMI) ≥30 kg/m2), were of western European origin, and aged between 30 and 70 years (N= 10,771).
5
clinical measures and definitions
Clinical measurements and laboratory methods
Detailed information about the physical examination and biochemical measurements has been published previously [22]. In short, during the first visit measurements of weight, waist circumference (WC), and height (to the nearest 0.5 cm) were performed in light clothing and without shoes. Body weight and height were used to calculate BMI (weight (kg)/height (m)2). Blood pressure (BP) was measured every minute during a period of 10 minutes with an automated DINAMAP Monitor (GE Healthcare, Freiburg, Germany). The average of the final three readings was recorded for systolic and diastolic BP. During the second visit, on average two weeks after the first visit, blood samples were drawn after an overnight fast for measurement of plasma glucose (hexokinase method), high density lipoprotein cholesterol (HDL-C) and triglycerides (TG) (respec-tively, colorimetric method and colorimetric UV method, Roche Modular P chemistry analyser, Basel, Switzerland).
Definition of the metabolically healthy and unhealthy obese
MHO was defined according to the criteria established by the BioSHaRE-EU Healthy Obesity Project [11], this means that subjects with obesity had none of the MetS risk factors, except for WC, according to the original NCEP ATPIII (including treatment for table 1. Definition for metabolically healthy obese (MHO) and metabolically unhealthy obese (MUO).
MHO Intermediate MUO
BMI ≥30 kg/m2 ≥30 kg/m2 ≥30 kg/m2
MetS risk factor none 1 risk factor ≥2 risk factors
Diagnosis for CVD no
MetS risk factor Threshold
Elevated blood pressure SBP ≥ 130 mmHg or
DBP ≥ 85 mmHg or use of antihypertensive medication
Impaired fasting glucose fasting blood glucose ≥ 6.1 mmol/L or
use of blood glucose lowering medication
or diagnosis of type 2 diabetes a
Decreased HDL-cholesterol b < 1.03 mmol/L in men or
< 1.30 mmol/L in women or medical treatment for low HDL-C
Elevated triglycerides b ≥ 1.70 mmol/L or medication for elevated
triglycerides
Abbreviations: BMI body mass index, CVD cardiovascular disease, DBP diastolic blood pressure, HDL high density lipoprotein cholesterol, SBP systolic blood pressure. a Diagnosis of type 2 diabetes was based on self-report and verified with self-reported medication use. b Subjects taking fibrates and/or nicotinic acid are presumed to have either high triglycerides and/or low HDL cholesterol.
dyslipidaemia, hyperglycaemia or hypertension) [23], and had no previous diagnosis of CVD (defined as self-reported myocardial infarction, stroke, or vascular intervention). MUO was defined as obesity with at least two MetS risk factors, while in ‘intermediate’ obesity only one MetS risk factor was present. Detailed information can be found in table 1.
dietary assessment
Food frequency questionnaire
We used a self-administered food frequency questionnaire (FFQ) to assess the habitual intake of 111 food items during the last month (4 weeks). After the first visit the FFQ was filled in by the participant at home and handed in, approximately 2 weeks later, at the LifeLines research centre during the second visit for fasting venepuncture. An existing validated Dutch FFQ formed the basis for the FFQ used in the LifeLines study [24, 25]. The basic FFQ focused on estimates of energy intake and macronutrients, including alcohol intake. For 46 main food items, frequency of consumption was indicated as ‘not this month’ or in days per week or month; including the amount (in units or specified portion size) consumed each time. The FFQ also included 37 questions on consumption of sub-items (e.g. 20+/30+ cheese, 40+ cheese, 48+ cheese, or cream cheese) for which fre-quency was specified as never, sometimes, often and (almost) always. Values of nutrient contents of foods were obtained from the 2006 Dutch food composition table (NEVO) [26].
To correct for potential under- or over-reporting on the dietary questionnaire, par-ticipants in the top and bottom 2.5% of daily energy intake (kcal/day) were excluded. In total, only 0.05% of the number of servings/food items data was missing, while frequency of consumption had been filled out. To calculate the participants’ intake, the average consumption of the food was used according to the Dutch National Food Consumption Survey 1998 [27]. Food items with missing data on frequency (0.5%) could not be interpreted and were not included in the calculations.
Dietary patterns for obese adults
Dietary patterns were derived on the basis of principal components analysis (PCA), a type of factor analysis. With PCA, linear combinations of the originally observed vari-ables are formed by grouping together correlated food items/groups, thus, identifying underlying components, or dietary patterns, within the data. The coefficients defining these linear combinations are called factor loadings and represent the correlations of each food item/group with the dietary pattern [28].
Since the proportion of explained variance per component (i.e. dietary pattern) decreases with the number of variables entered, individual food items with a similar
5
nutrient profile and culinary use were combined into 58 food groups (SupplementalTable S1). To test the appropriateness of applying PCA on the study sample, the Kaiser-Meyer-Olkin (KMO) measurement was conducted for testing sampling adequacy and Bartlett’s Test of Sphericity (BTS) was used to test the homogeneity of variances.
Next, the dietary patterns were derived on the basis of consumption (g/day) of each food group, unadjusted for energy intake. Within the PCA, orthogonal rotation (varimax option) was used to obtain uncorrelated patterns with greater interpretability. The deci-sion to retain a component was based on the following grounds: component Eigenvalue >1.0 (indicating that the component explains more of the variance in the correlations than is explained by a single variable), identification of an inflection point in the Scree plot, and interpretability of the pattern.
Stability of the derived components was assessed by comparing the components solutions and factor loadings in two random halves of the data set and per sex group. A component was considered stable if the same major patterns were identified, meaning the food groups with significant contributions (factor loading >0.3 or <-0.3) were similar.
A component score was created for each of the dietary patterns identified by multiplying the factor loadings by the corresponding standardized intake of the food (standardized for men and women separately), and summing across the food items/ groups for each pattern. These scores rank individuals according to the degree to which they adhered to the derived dietary pattern. Dietary patterns were named according to the foods with the highest loading on a component (considered as a loading >0.30).
physical activity
Physical activity was assessed by the validated SQUASH questionnaire (“Short QUes-tionnaire to ASsess Health-enhancing physical activity”) [29]. The SQUASH estimates habitual physical activities and is pre-structured in sports, commuting-, leisure time-, and household activities, and activities at work or school, referring to a normal week in the preceding months. Questions included type of activity, frequency, duration and in-tensity. Metabolic equivalent (MET) values were assigned to activities as defined by the Ainsworth’ compendium of Physical activities [30]. One MET unit is defined as the energy expenditure for sitting quietly. Activities with a MET value of 2 to < 4 were classified as light, 4 to < 6.5 as moderate, and ≥ 6.5 as vigorous intensity. A physical activity score was calculated by multiplying duration (minutes per week) with the MET value (taking into account the intensity at which an activity was performed). Subjects with implausible values from the SQUASH were excluded if: 1) ≥ 18 h/day was spent on activities listed in SQUASH [31], 2) separate categories exceeded plausible values, and/or 3) more than two activity categories of the questionnaire were missing.
demographic and lifestyle variables
Based on the participants’ responses to the self-administered questionnaires, data were assessed on the presence of diabetes mellitus, history of myocardial infarction, stroke or vascular intervention, current medication use, current use of a (self-)prescribed diet (e.g. energy-, fat- or salt restricted diet, prescribed diet for diabetes or high cholesterol, or fiber rich diet), education level, alcohol use and smoking. Education level was catego-rized as low (no formal education, only primary school or intermediate vocational edu-cation), medium (higher secondary education) or high (higher vocational education and university). Alcohol consumption was defined as non-drinker, <1-≤2 drinks/day (light-moderate) and >2 drinks/day (heavy) [22]. Smoking status was defined as non-smoker, former smoker and current smoker (including the use of cigarettes, cigarillos, cigars and pipe tobacco) [22]. Missing values on education level (0.6%) and smoking status (0.2%) were imputed, using single point imputation, with age, sex, netto household income and postal code (as a proxy for social economic status) as predictors.
Statistical analyses
All analyses were conducted using IBM SPSS Statistics version 22 (IBM Corporation, Armonk, NY,USA). All data are presented for men and women separately. Study charac-teristics were expressed as percentage (%), means with standard deviation (SD), or as median with interquartile range in case of non-normally distributed data. Differences between groups were tested by t-test for continuous variables or Kruskal-Wallis test when appropriate, and Chi-Square test for categorical variables. Multivariable logistic regression was used to determine the associations between MHO, dietary patterns (divided in quartiles) and physical activity scores (divided in tertiles) while adjusting for age and BMI (model 1), then smoking, alcohol use, education level, use of a diet and energy intake were added to the model (model 2). Subjects with MUO were used as the reference group. A P-value ≤0.05 was considered to be statistically significant.
rESultS
Study characteristics
After exclusion of participants with incomplete data on dietary intake, implausible physi-cal activity data, or missing data on cliniphysi-cal measures to define metabolic health, a total of 9,270 obese individuals were included in this study (86.1% from the original sample). Of those, 3,442 were men (37.1%) and 5,828 (62.9%) were women. The prevalence of MHO was 10.2% among men and 24.4% among women (Table 2). The prevalence of MUO was 56.9% in men and 35.3% in women, while the prevalence of ‘intermediate’ obesity was 32.9% in men and 40.3% in women. In general, MHO individuals had a
5
table 2. Clinical char ac ter istics ac cor ding t o metabolic health g roup . Men W omen MUO Int ermedia te MHO MUO Int ermedia te MHO N 1,959 (56.9) 1,131 (32.9) b 352 (10.2) b 2,058 (35.3) 2,348 (40.3) b 1,422 (24.4) b Age (y ears) 49.4 ± 9.4 49.6 ± 9.5 45.3 ± 8.6 b 50.6 ± 9.4 48.3 ± 9.9 b 44.9 ± 8.3 b BMI (k g/m 2) 32.2 [30.9-34.3] 31.8 [30.7-33.4] b 31.4 [30.7-32.9] b 33.6 [31.5-36.7] 32.8 [31.2-35.4] b 32.1 [30.9-34.4] b W aist cir cumf er enc e ( cm) 113.0 ± 9.0 111.1 ± 8.2 b 108.9 ± 7.2 b 108.0 ± 10.7 104.0 ± 9.6 b 101.4 ± 8.9 b Syst olic BP (mmHg) 139 ± 14 137 ± 14 b 122 ± 6 b 135 ± 15 131 ± 16 b 118 ± 7 b Diast olic BP (mmHg) 81 ± 9 81 ± 9 74 ± 5 b 77 ± 9 75 ± 9 b 70 ± 6 b Fasting gluc ose (mmol/L) 5.5 [5.1-6.2] 5.2 [5.0-5.6] b 5.2 [4.9-5.5] b 5.5 [5.0-6.3] 5.0 [4.7-5.4] b 4.9 [4.7-5.2] b HDL cholest er ol (mmol/L) 1.02 ± 0.21 1.28 ± 0.25 b 1.33 ± 0.20 b 1.20 ± 0.27 1.48 ± 0.33 b 1.58 ± 0.26 b Trigly cerides (mmol/L) 2.04 [1.60-2.72] 1.21 [0.96-1.50] b 1.08 [0.85-1.34] b 1.76 [1.25-2.21] 1.06 [0.82-1.35] b 0.91 [0.72-1.15] b Type 2 diabet es * (%) 10.5 0.6 b 0.0 13.3 0.8 b 0.0 CVD hist or y (%) 4.7 4.2 0.0 2.7 1.4 0.0 Use of BP lo wering dr ugs 33.0 20.4 b 0.0 44.4 24.7 b 0.0 Da ta ar e pr esen ted as mean ± SD or median [in ter quar tile range] or per cen tage (%). a denot es a P <0.01 and b denot es a P <0.0001 compar ed to MUO . BMI= body mass inde x; BP= blood pr essur e; HDL= high-densit y lipopr ot ein; CVD= car dio vascular disease . * Based on kno wn type 2 diabet es and newly -diag nosed type 2 diabet es (de-fined as a single fasting plasma gluc
ose lev
lower BMI and WC than MUO subjects (Table 2). The presence of the MHO phenotype decreased in older age groups for both men and women. Most men were defined as metabolically unhealthy obese, while most women below the age of 50 years were still defined as ‘intermediate’ (Figure 1).
Table 3 presents the macronutrient intake, lifestyle behaviors (physical activity, smoking and alcohol use) and education level across the metabolic health subgroups. No differ-ences in total energy intake or the energy-adjusted macronutrient intake were observed between the three metabolic health phenotypes. Obese men and women with MHO were more physically active than subjects with MUO. Furthermore, more men with MHO had a vigorous physical activity score in the highest tertile compared to men with MUO (43.8% vs 30.4%, P <0.0001). MHO women had more often a moderate physical activity score in the highest tertile compared to MUO women (36.1% vs 31.2%, P <0.01).
figure 1. Percentage of the metabolic health phenotype by age groups (left panel men, right panel
5
table 3. Macronutrient intake, lifestyle factors and education level according to metabolic healthpheno-type.
Men Women
MUO Intermediate MHO MUO Intermediate MHO
N 1,959 (56.9) 1,131 (32.9) b 352 (10.2) b 2,058 (35.3) 2,348 (40.3) b 1,422 (24.4) b
Energy intake (kcal/day) 2072 ± 488 2088 ± 496 2084 ± 492 1724 ± 443 1722 ± 451 1731 ± 465
Protein (% EI) 15.7 ± 2.3 15.6 ± 2.4 15.7 ± 2.5 16.4 ± 2.7 16.5 ± 2.7 16.4 ± 2.7
Plant protein (% EI) 6.1 ± 1.0 6.1 ± 1.0 6.2 ± 1.1 6.1 ± 1.0 6.1 ± 0.9 6.1 ± 1.0
Animal protein (% EI) 9.5 ± 2.5 9.4 ± 2.6 9.6 ± 2.8 10.4 ± 2.9 10.4 ± 2.8 10.3 ± 2.9
Carbohydrates (% EI) 45.0 ± 6.1 44.9 ± 6.4 45.2 ± 6.3 46.2 ± 6.3 46.0 ± 6.1 45.9 ± 6.2
Mono- and disaccharides (% EI) 19.6 ± 5.8 19.5 ± 5.5 19.7 ± 5.3 20.6 ± 5.8 20.6 ± 5.4 20.5 ± 5.2 Polysaccharides (% EI) 25.4 ± 4.8 25.3 ± 4.8 25.4 ± 4.7 25.5 ± 4.4 25.3 ± 4.5 25.3 ± 4.6 Fat (% EI) 36.0 ± 5.1 36.0 ± 5.1 36.0 ± 4.6 35.9 ± 5.4 35.9 ± 5.2 36.1 ± 5.1 Use of restricted/enriched diet (%) 5.1 4.9 6.6 12.6 12.2 10.6 MVPA (min/week) 285 [60-960] 360 [120-999] a 360 [140-1337] a 230 [60-600] 270 [100-646] a 242 [90-720] a Moderate PA score T1 ‘low’ 45.7 43.4 41.5 48.0 43.2 a 43.1 a T2 ‘medium’ 21.1 21.9 23.6 20.8 23.2 20.7 T3 ‘high’ 33.2 34.7 34.9 31.2 33.6 36.1 a Vigorous PA score T1 ‘low’ 36.1 31.2 a 25.9 a 36.2 33.6 32.5 T2 ‘medium’ 33.5 33.7 30.4 30.0 31.0 33.1 T3 ‘high’ 30.4 35.1 a 43.8 b 33.8 35.5 34.4 Smoking (%) Non 33.4 40.5 a 47.7 b 41.2 46.8 a 49.2 b Former 42.0 41.8 36.6 38.6 38.9 37.9 Current 24.6 17.7 b 15.6 a 20.2 14.3 b 12.9 b Use alcohol (%) Non 18.0 11.7 b 11.6 45.0 36.6 b 33.4 b ≤2 drinks 67.4 74.7 b 75.9 a 51.8 60.1 b 63.9 b >2 drinks 14.6 13.6 12.5 3.2 3.2 2.7 Education level (%) Low 39.7 39.4 36.4 46.8 40.1 b 32.2 b Medium 37.8 38.0 41.5 39.3 42.0 43.0 High 22.5 22.5 22.2 13.8 17.9 a 24.8 b
Data is presented as mean ± SD or median [interquartile range]. a denotes a P <0.01 and b denotes a P <0.0001. EI = energy intake; MUO= metabolically unhealthy obese; MHO= metabolically healthy obese; MVPA= moderate-vigorous physical active; PA= physical activity.
Since the metabolic health of a subject highly depends on age, we also checked the association between age and physical activity. Moderate physical activity scores were lower in older age groups, however, no clear association was found with the metabolic health status. Instead, vigorous physical activity scores were higher within older age groups, and higher among subjects with ‘intermediate’ obesity- and men with MHO compared to men with MUO (Figure 2).
figure 2. Moderate- and vigorous physical activity (PA) score (continuous) at the 75th percentile among
5
Fewer subjects with MHO were current smoker or did not use alcohol compared tosub-jects with MUO. The metabolically healthy obese men and women were more frequently light-moderate alcohol consumers (≤2 drinks/day). The distribution of education level was not different between the metabolic health subgroups in men, however, women with MHO were more frequent highly educated with 24.8% compared to 13.8% among women with MUO (P <0.0001).
dietary patterns in the obese population
To validate appropriateness of applying PCA on our study sample, we calculated the KMO and BTS values. The observed KMO was 0.73 suggesting that the study sample is suitable for PCA (should not be lower than 0.5). The BTS was significant (P <0.0001), indicating homogeneity of variance of the foods consumed. Figure 3 shows the Scree plot of Eigenvalues for each component. The Eigenvalues of the components dropped substantially until the fourth component (2.18). After the fourth component the Eigen-values remained more consistent, 1.65 for the fifth and 1.62 for the sixth component (Figure 3). As a result, we retained the 4 components solution. These four components overall explained 18.7% (5.7%, 4.5%, 4.4% and 4.1%, respectively) of the variations in food intake. PCA conducted on the two random halves of the dataset yielded similar results (data not shown). The same four major patterns were identified for men and women, although the magnitude of the loadings differed more as compared to the derived outcomes in the random halves of the dataset (data not shown).
Component Number 57 55 53 51 49 47 45 43 41 39 37 35 33 31 29 27 25 23 21 19 17 15 13 11 9 7 5 3 1 Eige nv alue 4 3 2 1 0 Scree Plot Page 1
The loadings of the food groups on the components (dietary patterns) are shown in Table 4. Positive loadings indicate that the subsequent food group is highly correlated with the corresponding dietary pattern, whereas negative loadings are inversely cor-related. The first dietary pattern we labeled as the ‘savory snacks and sweets’ pattern, which was characterized by high intakes of salty snackfood (warm and cold), fried pota-toes, sauces (warm sauces as well as mayonnaise and other non-red sauces) and sweets table 4. Loadings of the food groups on the dietary patterns.
Food group Savory snacks and Sweets
Meat and Alcohol Bread, Potatoes and Sweet snacks Vegetables, Fruit and Fish Warm sauces ,551 ,132 -,021 ,136 Savory snacks ,500 -,007 ,030 -,107 Fried potatoes ,470 ,128 ,058 -,186 Pasta ,459 ,151 ,027 ,188 Chocolate ,431 -,210 ,093 ,092 Cold sauces ,415 ,069 ,058 -,192
Commercially prepared dishes (ready to eat meals) ,398 ,032 -,189 -,111 Pastries ,384 -,087 ,313 -,037 Candybar ,383 -,134 ,082 -,075 Pizza ,377 ,022 -,124 -,127 Composed foods ,347 ,156 -,022 -,019 Rice ,343 ,142 -,051 ,297 Candy ,342 -,089 ,111 -,023 Salad dressing ,313 ,116 -,117 ,161
Peanuts, nuts and seeds ,292 ,163 -,006 ,057
Wipped cream ,267 ,027 ,120 ,012
Ice cream ,247 -,022 ,040 ,011
Low sugar bevareges ,219 -,008 -,016 -,065
Fruit juices ,134 ,002 ,025 -,093 Processed meat ,137 ,538 ,205 -,020 Beer ,017 ,434 -,086 -,287 Red meat -,004 ,429 ,214 -,047 Coffee -,072 ,415 ,012 -,109 Spirits ,055 ,389 -,138 -,118
Wine and fortified wine -,013 ,318 -,301 ,165
Eggs ,043 ,313 -,224 ,135
Lean red meat ,092 ,304 ,018 ,130
Legumes ,021 ,271 ,061 ,039
5
(chocolate, candy(bars) and pastries). The second pattern, labeled as the ‘meat and alcohol’ pattern, mainly consisted of processed meat, (lean) red meat, beer, spirits, wine/ fortified wine, eggs and coffee, but low consumption of tea and biscuits. Food groups in the third pattern, the so-called ‘bread, potatoes and sweet snacks’, had high loadings on bread, potatoes, edible fat, gravy, sweet sandwich toppings, pastries, biscuits and des-serts, whereas, wine/fortified wine and fish were inversely correlated with this pattern. The fourth dietary pattern labeled as ‘fruit, vegetables and fish’ was characterized by high table 4. Loadings of the food groups on the dietary patterns. (continued)
Food group Savory snacks and Sweets
Meat and Alcohol Bread, Potatoes and Sweet snacks Vegetables, Fruit and Fish Soup ,059 ,198 -,040 ,072 Beer – light -,002 ,118 -,039 -,048 Bread ,015 ,302 ,599 -,037 Edible fat ,010 ,299 ,591 -,045 Potatoes -,099 ,304 ,570 ,015
Sweet sandwich toppings ,094 -,185 ,498 ,077
Gravy -,049 ,340 ,486 -,076
Biscuits ,261 -,300 ,328 ,217
Desserts -,024 -,024 ,301 ,028
Apple sauce ,067 -,085 ,254 -,121
Nonfermented medium /low fat milk ,040 -,022 ,172 ,002
Fermented milk products - sweetened ,061 -,093 ,136 ,021
Vegetables -,002 ,238 ,118 ,543
Fruit -,182 -,073 ,091 ,454
Warm savory snacks ,417 ,183 ,056 -,434
High sugar beverages ,231 ,046 ,074 -,396
Tea ,025 -,306 ,093 ,391
Fatty fish ,005 ,187 -,293 ,357
Mayonnaise ,326 ,177 -,014 -,338
Lean fish ,057 ,154 -,267 ,336
Fermented milk products - unsweetened -,085 -,010 ,060 ,322
Added sugar ,011 ,113 ,154 -,295
Chicken ,107 ,169 -,052 ,278
Quark -,003 -,066 -,056 ,239
Cheese – low fat -,173 ,030 ,080 ,237
Cereals ,104 -,104 -,033 ,209
Chocolate milk ,107 -,096 ,124 -,164
Nonfermented whole milk -,039 ,092 ,054 -,119
figur e 4. Dietar y pa tt er n sc or es acr oss age g roups (lef t panel men, r igh t panel w omen).
5
consumption of fruit, vegetables, fatty and lean fish, tea and unsweetened fermentedmilk products, and low consumption of high sugar beverages, mayonnaise and salty snackfood.
dietary patterns and demographic- and lifestyle characteristics
Because dietary patterns are part of a broader pattern of lifestyle factors, demographic and lifestyle characteristics are shown by quartiles of dietary pattern scores in Table 5. Di-etary pattern scores were also checked across age groups (Figure 4). Interpreting Table 5 and Figure 4 combined, we came to the following observations. Subjects in the highest quartile of the ‘savory snacks and sweets’ pattern, were of a younger age, more often metabolically healthy obese, higher educated and had less often a chronic condition (T2D, history of CVD, or hypertension) or used a (self-) prescribed diet. At the same time, they were less physically active and more often smokers. Scores on both the ‘meat and alcohol’ pattern and the ‘bread, potatoes and sweet snacks’ were stable across age groups, but subjects with a high score on these patterns were less often metabolically healthy obese and were less educated compared to subjects with a low score. The ‘meat and alcohol’ pattern was, furthermore, characterized by less physical activity, smoking and heavy alcohol use (17.4% consuming >2 drinks/day). In contrast, subjects in the highest quartile of the ‘bread, potatoes and sweet snacks’ pattern were more physically active, were less often a smoker and drank less alcohol. Older people were more likely to fit the ‘fruit, vegetables and fish’ pattern. The pattern was also associated with higher education levels, more physical activity, less smokers (although with more former smok-ers) and less alcohol use. However, subjects with a high score on the ‘ fruit, vegetables and fish’ pattern were more likely to have a chronic condition and used more often a (self-) prescribed diet.
dietary patterns and physical activity as determinants of mho
Table 6 shows the multivariable associations between lifestyle behaviors and MHO in men and women, taking into account differences in age, BMI, education level, following a (self-) prescribed diet and energy intake. There was a gradual lower odds ratio (OR) for the MHO phenotype within higher age groups. Also, men and women with a BMI ≥35 kg/m2 were less likely MHO.
Of the four obesity-specific dietary patterns, two patterns were associated with MHO in women only. Among women, a higher score on the ‘fruit, vegetables and fish’ pattern, non-smoking and alcohol consumption up to 2 drinks/day was associated with a higher OR for MHO. However, higher scores on the ‘bread, potatoes and sweet snacks’ pattern was associated with a lower OR for the MHO phenotype. Among men, higher vigorous
table 5.
D
emog
raphic and lif
est
yle v
ar
iables within the first and f
our
th quar
tile of the dietar
y pa tt er n sc or es . Sa vor y snacks and sw eets Mea t and alcohol Br ead , pota toes and sw eet snacks Fruit , v
egetables and fish
Q1 Q4 Q1 Q4 Q1 Q4 Q1 Q4 N 2,317 2,317 2,317 2,317 2,317 2,317 2,317 2,317 Males (%) 860 (37.1) 860 (37.1) 860 (37.1) 860 (37.1) 860 (37.1) 860 (37.1) 860 (37.1) 859 (37.1) Age (y ears), mean ± sd 53.9 ± 9.9 44.2 ± 7.4 b 47.9 ± 10.1 48.7 ± 8.9 49.0 ± 9.5 48.4 ± 9.8 44.8 ± 8.1 51.6 ± 9.9 b BMI (k g/m 2), median [IQR] 33.5 [30.9-34.7] 32.6 [31.1-35.3] a 32.5 [30.9-34.8] 32.5 [31.0-35.2] 32.4 [30.9-34.8] 32.4 [30.9-34.9] 32.7 [31.1-35.4] 32.3 [30.9-34.6] b
Metabolic health (%) MUO
46.4 40.1 b 42.1 43.8 39.4 44.9 a 44.8 41.3 In termedia te 38.6 37.5 36.3 37.9 38.5 37.4 36.7 38.9 MHO 15.0 22.4 b 21.7 18.3 a 22.1 17.7 a 18.5 19.8 Ener gy in tak e (k cal/da y) 1486 ± 412 2279 ± 411 b 1572 ± 449 2191 ± 456 b 1504 ± 432 2256 ± 404 b 1973 ± 511 1813 ± 490 b MVP A (min/w eek) 280 [110-720] 255 [90-769] 268 [90-720] 280 [90-780] 250 [75-660] 300 [100-840] b 270 [60-960] 290 [120-680] Moder at e P A sc or e 225 [0-1350] 300 [0-2625] a 240 [0-2025] 300 [0-1700] a 300 [0-1700] 300 [0-2700] a 300 [0-3780] 300 [0-1500] a Vigor ous P A sc or e 960 [105-2160] 645 [0-1600] b 800 [48-1920] 640 [0-1680] b 720 [0-1800] 780 [80-1920] a 480 [0-1440] 960 [240-2160] b Educ ation (%) Lo w 52.1 29.8 b 37.1 41.4 a 32.2 46.9 b 47.5 32.8 b Medium 33.4 45.7 b 41.1 40.5 40.1 39.4 41.1 38.5 High 14.5 24.5 b 21.8 18.1 a 27.7 13.8 b 11.4 28.6 b Smok ing (%) Non 38.6 46.3 b 54.4 31.3 b 38.6 46.6 b 39.0 44.2 a Former 45.0 34.0 b 34.9 42.8 b 42.4 37.5 a 30.3 46.7 b Curr en t 16.4 19.7 a 10.7 25.9 b 19.0 15.9 a 30.7 9.1 b
5
table 5.
D
emog
raphic and lif
est
yle v
ar
iables within the first and f
our
th quar
tile of the dietar
y pa tt er n sc or es . (c on tinued) Sa vor y snacks and sw eets Mea t and alcohol Br ead , pota toes and sw eet snacks Fruit , v
egetables and fish
Q1 Q4 Q1 Q4 Q1 Q4 Q1 Q4 Use alc ohol (%) Non 33.0 29.0 a 44.1 17.9 b 20.3 38.8 b 35.5 25.6 b ≤2 drinks 60.2 63.2 55.0 64.7 b 68.3 57.1 b 52.9 69.1 b >3 drinks 6.8 7.8 0.9 17.4 b 11.4 4.1 b 11.6 5.4 b (S elf -) pr escribed diet (%) 18.2 3.9 b 11.8 7.0 b 15.8 5.2 b 3.5 18.9 b Type 2 diabet es (%) 10.3 2.8 b 5.6 5.0 4.5 5.4 3.2 6.6 b CVD hist or y (%) 4.4 1.2 b 2.2 2.4 2.6 2.8 1.8 3.4 a Use of BP lo wering dr ug (%) 37.3 17.7 b 24.3 25.2 25.2 24.5 19.5 29.4 b Da ta is pr esen ted as mean ± SD or median [in ter quar tile range] or per cen tage (%). a denot es a P <0.01 and b denot es a P <0.0001 compar ed to Q1 of the same dietar y pa tt er n. MVP A= moder at e– vigor ous ph ysical ac tivit y; C VD= car dio
vascular disease; BP= blood pr
essur
PA, non-smoking and alcohol consumption was associated with a higher OR for MHO (model 2, Table 6).
Since we have adapted the revised NCEP ATPIII by using a less strict threshold for impaired fasting glucose (≥6.1 mmol/L), in a subsequent analysis we applied the more strict cut-off of ≥5.6 mmol/L to define metabolic health (data not shown). This resulted in 196 fewer individuals classified as MHO and 566 more individuals classified as MUO. The observed associations in the multivariable regression were stronger (generally first decimal place increased), for both men and women, using the more strict threshold. No new associations were found.
table 6. Multivariable-adjusted odds ratios for the associations of demographic- and lifestyle factors with
the metabolically healthy obesity phenotype.
Men Women
Model 1 Model 2 Model 1 Model 2 Age group 30-39 1 1 1 1 40-49 0.75 (0.56-1.01) 0.75 (0.56-1.01) 0.45 (0.37-0.55) 0.47 (0.38-0.57) 50-59 0.36 (0.25-0.53) b 0.36 (0.24-0.53) b 0.22 (0.17-0.28) b 0.22 (0.17-0.28) b 60-69 0.33 (0.21-0.53) b 0.33 (0.21-0.54) b 0.10 (0.07-0.13) b 0.09 (0.07-0.13) b BMI group 30-34.9 1 1 1 ≥35 0.37 (0.24-0.55) b 0.34 (0.23-0.52) b 0.38 (0.32-0.45) b 0.39 (0.33-0.46) b Dietary pattern
Savory snacks and sweets
Q1 1 1 1 1
Q2 1.12 (0.78-1.63) 1.16 (0.79-1.71) 1.26 (1.01-1.57) 1.14 (0.90-1.44)
Q3 1.46 (1.02-2.09) 1.55 (1.03-2.33) 1.13 (0.90-1.42) 0.99 (0.77-1.28)
Q4 1.29 (0.88-1.90) 1.43 (0.87-2.36) 1.43 (1.13-1.81) a 1.27 (0.92-1.75)
Meat and alcohol
Q1 1 1 1 1
Q2 0.89 (0.64-1.23) 0.90 (0.64-1.26) 0.98 (0.80-1.21) 1.00 (0.81-1.24)
Q3 0.78 (0.56-1.09) 0.83 (0.58-1.20) 0.83 (0.67-1.03) 0.81 (0.65-1.01)
Q4 0.72 (0.51-1.02) 0.85 (0.56-1.29) 1.02 (0.82-1.26) 1.02 (0.80-1.31)
Bread, potatoes and sweet snacks
Q1 1 1 1 1
Q2 0.78 (0.56-1.09) 0.79 (0.56-1.12) 0.64 (0.52-0.80) 0.68 (0.55-0.85) a
Q3 0.75 (0.54-1.06) 0.79 (0.54-1.16) 0.66 (0.53-0.81) 0.72 (0.56-0.92) a
5
table 6. Multivariable-adjusted odds ratios for the associations of demographic- and lifestyle factors withthe metabolically healthy obesity phenotype. (continued)
Men Women
Model 1 Model 2 Model 1 Model 2
Fruit, vegetables and fish
Q1 1 1 1 1 Q2 1.03 (0.75-1.44) 1.01 (0.72-1.41) 1.24 (1.01-1.54) 1.11 (0.89-1.38) Q3 0.88 (0.62-1.24) 0.81 (0.57-1.16) 1.55 (1.25-1.92) b 1.31 (1.04-1.64) Q4 0.96 (0.68-1.36) 0.87 (0.60-1.27) 1.75 (1.40-2.19) b 1.46 (1.15-1.87) a Moderate PA score T1 ‘low’ 1 1 1 1 T2 ‘medium’ 1.22 (0.90-1.65) 1.17 (0.86-1.59) 1.10 (0.90-1.34) 1.07 (0.87-1.30) T3 ‘high’ 1.03 (0.78-1.35) 0.98 (0.73-1.30) 1.15 (0.97-1.36) 1.17 (0.98-1.38) Vigorous PA score T1 ‘low’ 1 1 1 1 T2 ‘medium’ 1.21 (0.89-1.65) 1.18 (0.86-1.61) 1.17 (0.97-1.40) 1.13 (0.94-1.35) T3 ‘high’ 2.11 (1.57-2.83) b 1.96 (1.45-2.64) b 1.21 (1.01-1.45) 1.17 (0.97-1.40) Smoking Non 1 1 Former 0.73 (0.55-0.96) 0.93 (0.78-1.10) Current 0.44 (0.31-0.63) b 0.49 (0.39-0.62) b Alcohol Non 1 1 ≤ 2 drink/day 1.86 (1.28-2.70) a 1.59 (1.34-1.88) b > 2 drinks/day 1.81 (1.07-3.04) 1.33 (0.82-2.15) Education Low 1 1 Middle 1.00 (0.76-1.31) 0.94 (0.79-1.12) High 0.86 (0.61-1.21) 1.22 (0.97-1.53) Use of diet No 1 1 Yes 1.46 (0.88-2.44) 0.71 (0.56-0.92) a Energy intake* 0.81 (0.51-1.29) 0.93 (0.67-1.28) Data are expressed as odds ratios (95% confidence interval). Reference group is the metabolically unhealthy obese. Values in BOLD indicate a P <0.05, a denotes a P <0.01 and b denotes a P <0.0001. Dietary patterns are presented in quartiles, and moderate- and vigorous physical activity (PA) score are presented in tertiles. * Energy intake is given per 1.000 kcal increase.
diScuSSioN
In this study of 9,270 obese adults between 30-70 years of age, more than half of the obese men and more than one third of obese women were metabolically unhealthy. Only 10% of men, and 25% of women were metabolically healthy obese. We found that compared to women with MUO women with MHO had a healthier diet, rich in fruit, vegetables, fish and fermented milk products (unsweeted) while avoiding high sugar beverages and, salty- and sweet snackfood. Compared to the metabolically unhealthy obese, men with MHO were characterized by higher engagement in intensive vigorous physical activity. In addition, moderate alcohol consumption and non-smoking were positively associated with MHO in both men and women.
dietary patterns as determinant of mho
Dietary and lifestyle factors are known to play an important role in the development of insulin resistance, obesity, metabolic syndrome and T2D [32, 33]. Consistent with previous data, in our obese population total energy intake and dietary macronutrient composition did not differ between subjects with MHO and subjects with MUO [13-16]. We identified four major obesity-specific dietary patterns, which we called the ‘savory snacks and sweets’ pattern, the ‘meat and alcohol’ pattern, the ‘bread, potatoes and sweet snacks’ pattern and the ‘fruit, vegetables and fish’ pattern. Higher scores on the ‘fruit, veg-etables and fish’ pattern were positively associated with MHO in a dose-dependent way, whereas higher scores on the ‘bread, potatoes and sweet snacks’ pattern were inversely associated with MHO. These associations were only observed in women.
The ‘fruit, vegetables and fish’ pattern mainly consisted of foods, which we often consider as healthy. Previous studies showed that higher intakes of vegetables and fruit were associated with a lower risk of the MetS and CVD [34], and a higher fruit consump-tion was associated with lower risk of T2D [35, 36]. Furthermore, higher scores on the ‘fruit, vegetables and fish’ pattern mean higher intake of fish, chicken, fermented milk products (unsweetened) and low-fat cheese, which suggests a higher overall protein intake. Both epidemiological and experimental data show that the consumption of dairy products have a beneficial effect on MetS risk factors and are associated with a lower risk of body fat gain and obesity as well as CVD [37]. Consumption of cheese and fermented dairy product were also inversely associated with T2D incidence [38, 39].
The pattern of ‘bread, potatoes and sweet snacks’ was inversely associated with MHO in our study. This pattern reflects a diet high in carbohydrates. Bread was mainly eaten in combination with sweet sandwich toppings, a typical Dutch eating habit. Furthermore, mainly sweet snacks were consumed like pastries, biscuits and desserts. These products are often labeled as foods with a high glycaemic index (GI), causing a quick, but only short-term supply of energy [40]. Although mixed results were reported for the
expe-5
rienced feeling of hunger and energy-intake after such high GI food or meal [41, 42],intervention studies have reported an inverse association between high GI diets and HDL-C [43, 44] and a positive association with triglycerides [45]. Furthermore, a higher GI was associated with the development of MetS [46]. Our observed finding is in line with data that suggest that high GI foods may deteriorate the metabolic risk profile. In practice, for those obese individuals who attempt to improve their health but experi-ence difficulties to lose weight, shifting to the ‘fruit, vegetables and fish’-enriched diet may be more appealing or easier option to consider.
To date, only a few studies examined dietary patterns and metabolic health in obesity. In two studies ‘a priori’ dietary scores were applied [16, 19]. A study in obese adolescents and adults in the US used the Healthy Eating Index (HEI-2005) and found that the total score was higher in MHO adolescents and adult women compared with metabolically abnormal obese, e.g. MUO [19]. In an Irish cohort study, there was no asso-ciation between dietary quality (the extent to which the eating behaviour is “healthy”), based on the DASH score (Dietary Approaches to Stop Hypertension)), and MHO [16]. The food pyramid score, representing the optimal number of servings from each basic food groups in daily food intake, was higher among MHO. However, in adjusted analysis (for sex, age, physical activity, alcohol, smoking and dietary quality) it was no longer as-sociated with metabolic health. To the best of our knowledge there is only one study by Bell et al. [47] which used the ‘a posteriori’ approach, where actual dietary patterns were determined, which are habitual in nature [18]. The odds of having a more metabolically healthy profile was 16% greater for every standard deviation increase in the ‘healthy’ dietary pattern score (including high loadings of whole grains, fresh fruit, dried fruit, legumes and low fat dairy) [47]. Although the criteria for metabolic health between our study and the one by Bell et al. [47] were different and dietary patterns were based on a population including non-obese and obese participants, based on the outcomes of both studies we hypothesize that small changes in the diet may improve metabolic health risk.
Our findings for dietary patterns and MHO were observed for women only. Such a sex difference may reflect differences in physiology, reporting of diet, or the amount consumed of the specific types of foods that contributed strongly to the pattern score [48]. For instance, women reported higher consumptions of fruit, vegetables, fish and fermented milk products (unsweetened) which contribute highly to the ‘fruit, vegetables and fish’ pattern, whereas pastries, biscuits and desserts which contribute high to the ‘bread, potatoes and sweet snacks’ pattern, were also more consumed by women than men (data not shown). Thus, higher pattern scores in women did also mean higher intakes of the specific foods likely to be associated with metabolic (un)health. We also hypothesize that men and women have a different view on improving health status.
While women will mainly adapt their diet to lose weight, men are more focused on improving their physical fitness.
The role demographic- and lifestyle factors in dietary patterns
In addition to the observed sex difference in eating behaviour, age played an important role. Subjects with a high score on the ‘savory snacks and sweets’ pattern were younger and had fewer chronic conditions, whereas subjects with a high score on the ‘fruit, vegetables and fish’ pattern were older and had more chronic conditions. While the first pattern is intuitively seen as an unhealthy dietary pattern, the latter reflects more health consciousness. Health status may, therefore, be important for the food choices made. Dietary habits may also be a proxy for other (lifestyle) variables, such as physical activity, smoking, alcohol use and educational attainment. This suggests that it is necessary to take a more holistic approach when dietary patterns are studied, otherwise incorrect conclusions could be drawn upon the possible effects of foods.
physical activity as determinant of mho
Increasing one’s physical activity has the potential to improve adiposity profile and metabolic risk, even in the absence of weight loss [49]. A more favourable fat distribu-tion, with less visceral fat, was associated with a long-term metabolically healthy profile in obese adults over a period of 10 year and no excess risk of T2D and CVD [50].
In line with our results, other studies on physical activity and MHO found that both objectively measured total physical activity [12] and self-reported moderate-vigorous physical activity were higher in the MHO group compared to MUO group [51-53]. In the present study, especially the level of vigorous physical activity was an important feature in the relationship between physical activity and MHO, found only in men. In general, men engage in more physical activity compared to women (data not shown). We, there-fore, hypothesize that this results in a stronger positive association between vigorous physical activity and MHO in men. Another explanation may come from the potential of biased results. SQUASH is a self-reported physical activity questionnaire, where subjects are asked to specify the intensity level of each performed activity. Women may experi-ence certain activities as more intensive than men, while actually a lower cardiorespira-tory response is obtained for the same type of physical activity and the same indicated intensity level. As a result men may develop a higher cardiorespiratory fitness level than women, which has been linked to MHO [54].
Epidemiological observations provide new indications that it is important to reduce time in sedentary behaviors (e.g. activities involving low levels of metabolic energy ex-penditure, primarily sitting and laying down) [55]. In a recent study, metabolically healthy obese had a higher total step count and were on average less sedentary compared to metabolically unhealthy obese [56]. Studies on the interactive nature of physical
activ-5
ity and sitting time on metabolic risk factor clustering were inconclusive. While somestudies suggest that the strength of the association between sitting and metabolic risk depends on the engagement in physical activity [57, 58], other studies reported that higher sitting time was associated with metabolic risk independently of physical activity [59-61]. Since the SQUASH questionnaire, used in LifeLines, is not designed to capture sedentary behavior, we were not able to compare its levels among the metabolic health phenotypes.
Strengths and limitations
Our study includes a representative sample of the Dutch population using extensive questionnaires to measure important lifestyle behaviors, and standardized protocols to obtain clinical and biochemical measurements. Another strength of our study is that it is the largest study on this topic to date and the use of obesity-specific dietary patterns.
However, there are also some limitations. Dietary intake, physical activity and other lifestyle behaviors were based on self-reported data, and are subject to recall bias. Obese individuals tend to underestimate their dietary intake and overestimate their physical activity [62]. However, in our study we only used obese subjects to make comparisons, hence, over- and underreporting are likely to be equally distributed over the three metabolic health groups. Furthermore, we ranked individuals into categories, rather than using the estimated absolute quantities, which reduces the effect of over- or underreporting.
Although PCA is extensively used in nutritional epidemiology and showed reason-able reproducibility and validity using FFQ data [63-66], more validation studies are needed. Furthermore, PCA requires several arbitrary decisions, such as the pre-selection of food groups, the number of retained patterns, the method of rotation, and the cut-off value used to define a significant contribution of the factor loadings [28]. Yet, it was found that derived dietary patterns were robust for subjective factor analytical decisions [67].
The four dietary patterns explained 18.7% of the total variance in consumption of the food groups. Although this seems low compared to other studies using PCA, the percent variance explained is a function of the number of food items included in the fac-tor analysis [28]. We used more food items compared to others [48, 68, 69], which results in a lower explained variance, though, the derived patterns are more detailed [28].
Next, we utilized cross-sectional data from which we cannot infer causality. It is, therefore, not possible to rule out reverse causality. Individuals which experience nega-tive health outcomes, such as having T2D, a CVD history or the knowledge of having dyslipidaemia or hypertension, may have changed their dietary intake and their physical activity. We consider it likely that dietary patterns are not isolated, but part of a broader lifestyle and subject to someone’s phase in life.
coNcluSioN
Our research showed that key lifestyle behaviors differed between metabolically healthy obese and metabolically unhealthy obese adults aged 30-69 years. While non-modifiable factors like age and sex are important for determining someone’s baseline odds for MHO, our data suggests that a healthier diet in women and vigorous physical activity in men (as well as non-smoking and moderate alcohol use) may be related to this favourable obesity state. Identification of behavioural lifestyle patterns may help in pinpointing vulnerable subgroups in the obese population and to develop potential strategies improving metabolic health.
acKNowlEdgEmENtS
The authors wish to acknowledge the services of the LifeLines Cohort Study, the contrib-uting research centers delivering data to LifeLines, and all the study participants.
5
rEfErENcES
1. Swinburn BA, Sacks G, Hall KD, McPherson K, Finegood DT, Moodie ML, Gortmaker SL: The global obesity pandemic: shaped by global drivers and local environ-ments. Lancet (London, England) 2011, 378(9793):804-814.
2. Mokdad AH, Ford ES, Bowman BA, Dietz WH, Vinicor F, Bales VS, Marks JS: Prevalence of obesity, diabetes, and obesity-related health risk factors, 2001. JAMA 2003, 289(1):76-79.
3. Denis GV, Obin MS: ‘Metabolically healthy obesity’: origins and implications. Molecular aspects of medicine 2013, 34(1):59-70.
4. Bell JA, Kivimaki M, Hamer M: Metabolically healthy obesity and risk of incident type 2 diabetes: a meta-analysis of prospective cohort studies. Obesity reviews :
an official journal of the International Association for the Study of Obesity 2014, 15(6):504-515.
5. Kramer CK, Zinman B, Retnakaran R: Are metabolically healthy overweight and obesity benign conditions?: A systematic review and meta-analysis. Annals of internal
medicine 2013, 159(11):758-769.
6. Ebbeling CB, Swain JF, Feldman HA, Wong WW, Hachey DL, Garcia-Lago E, Ludwig DS: Effects of dietary composition on energy expenditure during weight-loss maintenance. JAMA 2012, 307(24):2627-2634. 7. Kraschnewski JL, Boan J, Esposito J, Sherwood NE,
Lehman EB, Kephart DK, Sciamanna CN: Long-term weight loss maintenance in the United States.
Inter-national journal of obesity (2005) 2010,
34(11):1644-1654.
8. Achilike I, Hazuda HP, Fowler SP, Aung K, Lorenzo C: Predicting the development of the metabolically healthy obese phenotype. International journal of
obesity (2005) 2015, 39(2):228-234.
9. Eshtiaghi R, Keihani S, Hosseinpanah F, Barzin M, Azizi F: Natural course of metabolically healthy abdominal obese adults after 10 years of follow-up: the Tehran Lipid and Glucose Study. International journal of obesity
(2005) 2015, 39(3):514-519.
10. Hamer M, Bell JA, Sabia S, Batty GD, Kivimaki M: Stabil-ity of metabolically healthy obesStabil-ity over 8 years: the English Longitudinal Study of Ageing. European journal
of endocrinology / European Federation of Endocrine Societies 2015, 173(5):703-708.
11. van Vliet-Ostaptchouk JV, Nuotio ML, Slagter SN, Doiron D, Fischer K, Foco L, Gaye A, Gogele M, Heier M, Hiekkalinna T et al: The prevalence of metabolic syndrome and metabolically healthy obesity in Europe: a collaborative analysis of ten large cohort studies. BMC
endocrine disorders 2014, 14:9.
12. Bell JA, Hamer M, van Hees VT, Singh-Manoux A, Kivi-maki M, Sabia S: Healthy obesity and objective physical activity. The American journal of clinical nutrition 2015, 102(2):268-275.
13. Hankinson AL, Daviglus ML, Van Horn L, Chan Q, Brown I, Holmes E, Elliott P, Stamler J: Diet composition and activity level of at risk and metabolically healthy obese American adults. Obesity (Silver Spring, Md) 2013, 21(3):637-643.
14. Kimokoti RW, Judd SE, Shikany JM, Newby PK: Food intake does not differ between obese women who are metabolically healthy or abnormal. The Journal of
nutrition 2014, 144(12):2018-2026.
15. Kimokoti RW, Judd SE, Shikany JM, Newby PK: Meta-bolically Healthy Obesity Is Not Associated with Food Intake in White or Black Men. The Journal of nutrition 2015, 145(11):2551-2561.
16. Phillips CM, Dillon C, Harrington JM, McCarthy VJ, Kear-ney PM, Fitzgerald AP, Perry IJ: Defining metabolically healthy obesity: role of dietary and lifestyle factors.
PloS one 2013, 8(10):e76188.
17. Hu FB: Dietary pattern analysis: a new direction in nutritional epidemiology. Current opinion in lipidology 2002, 13(1):3-9.
18. Michels KB, Schulze MB: Can dietary patterns help us detect diet-disease associations? Nutrition research
reviews 2005, 18(2):241-248.
19. Camhi SM, Whitney Evans E, Hayman LL, Lichtenstein AH, Must A: Healthy eating index and metabolically healthy obesity in U.S. adolescents and adults.
Preven-tive medicine 2015, 77:23-27.
20. Klijs B, Scholtens S, Mandemakers JJ, Snieder H, Stolk RP, Smidt N: Representativeness of the LifeLines Cohort Study. PloS one 2015, 10(9):e0137203.
21. Scholtens S, Smidt N, Swertz MA, Bakker SJ, Dotinga A, Vonk JM, van Dijk F, van Zon SK, Wijmenga C, Wolffen-buttel BH et al: Cohort Profile: LifeLines, a
three-generation cohort study and biobank. International
journal of epidemiology 2015, 44(4):1172-1180.
22. Slagter SN, van Vliet-Ostaptchouk JV, Vonk JM, Boezen HM, Dullaart RP, Kobold AC, Feskens EJ, van Beek AP, van der Klauw MM, Wolffenbuttel BH: Combined effects of smoking and alcohol on metabolic syndrome: the LifeLines cohort study. PloS one 2014, 9(4):e96406. 23. Expert Panel on Detection Evaluation, and Treatment of
High Blood Cholesterol in Adults: Executive Summary of The Third Report of The National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, And Treatment of High Blood Cholesterol In Adults (Adult Treatment Panel III). JAMA 2001, 285(19):2486-2497.
24. Siebelink E, Geelen A, de Vries JH: Self-reported energy intake by FFQ compared with actual energy intake to maintain body weight in 516 adults. The British journal
of nutrition 2011, 106(2):274-281.
25. Streppel MT, de Vries JH, Meijboom S, Beekman M, de Craen AJ, Slagboom PE, Feskens EJ: Relative validity of the food frequency questionnaire used to assess dietary intake in the Leiden Longevity Study. Nutrition journal 2013, 12:75.
26. Bureau stichting NEVO, NEVO-tabel 2006. In .: Den Haag: Voedingscentrum; 2006.
27. Zo eet Nederland: resultaten van de Voedselconsump-tiepeiling 1997-1998. In .: Den Haag: Voedingscentrum; 1998.
28. Newby PK, Tucker KL: Empirically derived eating pat-terns using factor or cluster analysis: a review. Nutrition
reviews 2004, 62(5):177-203.
29. Wendel-Vos GC, Schuit AJ, Saris WH, Kromhout D: Reproducibility and relative validity of the short questionnaire to assess health-enhancing physical activity. Journal of clinical epidemiology 2003, 56(12):1163-1169.
30. Ainsworth BE, Haskell WL, Herrmann SD, Meckes N, Bassett DR, Jr., Tudor-Locke C, Greer JL, Vezina J, Whitt-Glover MC, Leon AS: 2011 Compendium of Physical Activities: a second update of codes and MET values. Medicine and science in sports and exercise 2011, 43(8):1575-1581.
31. Sijtsma A, Sauer PJ, Corpeleijn E: Parental correlations of physical activity and body mass index in young children--he GECKO Drenthe cohort. The international
journal of behavioral nutrition and physical activity
2015, 12:132.
32. Harrington JM, Phillips CM: Nutrigenetics: bridging two worlds to understand type 2 diabetes. Current diabetes
reports 2014, 14(4):477.
33. Phillips CM: Nutrigenetics and metabolic disease: cur-rent status and implications for personalised nutrition.
Nutrients 2013, 5(1):32-57.
34. Esmaillzadeh A, Kimiagar M, Mehrabi Y, Azadbakht L, Hu FB, Willett WC: Fruit and vegetable intakes, C-reactive protein, and the metabolic syndrome. The
American journal of clinical nutrition 2006,
84(6):1489-1497.
35. Bazzano LA, Li TY, Joshipura KJ, Hu FB: Intake of fruit, vegetables, and fruit juices and risk of diabetes in women. Diabetes care 2008, 31(7):1311-1317. 36. Muraki I, Imamura F, Manson JE, Hu FB, Willett WC,
van Dam RM, Sun Q: Fruit consumption and risk of type 2 diabetes: results from three prospective longi-tudinal cohort studies. BMJ (Clinical research ed) 2013, 347:f5001.
37. Astrup A: Yogurt and dairy product consumption to prevent cardiometabolic diseases: epidemiologic and experimental studies. The American journal of clinical
nutrition 2014, 99(5 Suppl):1235s-1242s.
38. O’Connor LM, Lentjes MA, Luben RN, Khaw KT, Wareham NJ, Forouhi NG: Dietary dairy product intake and incident type 2 diabetes: a prospective study using dietary data from a 7-day food diary. Diabetologia 2014, 57(5):909-917.
39. Sluijs I, Forouhi NG, Beulens JW, van der Schouw YT, Agnoli C, Arriola L, Balkau B, Barricarte A, Boeing H, Bueno-de-Mesquita HB et al: The amount and type of dairy product intake and incident type 2 diabetes: results from the EPIC-InterAct Study. The American
journal of clinical nutrition 2012, 96(2):382-390.
40. Ludwig DS: The glycemic index: physiological mecha-nisms relating to obesity, diabetes, and cardiovascular disease. JAMA 2002, 287(18):2414-2423.
41. Bornet FR, Jardy-Gennetier AE, Jacquet N, Stowell J: Glycaemic response to foods: impact on satiety and long-term weight regulation. Appetite 2007, 49(3):535-553.
42. Sun FH, Li C, Zhang YJ, Wong SH, Wang L: Effect of Glycemic Index of Breakfast on Energy Intake at
