Journal of Obesity & Weight Loss Therapy
Open Access

BMI: Body Mass Index; CI: Confidence Interval; NOW: Normal-Weight Obesity; OR: Odds Ratio

Weight and body mass index (BMI) are increasing in the Western world [1]. Increasing BMI is associated with un-health and higher prevalence of metabolic diseases [2,3]. Not only weight and BMI may be of interest for health, but also the body constitution estimated as fat percentage and waist/hip ratio, which are less seldom studied [4,5]. There is a lot of debate and controversy about which lifestyle habits are most efficient to maintain a normal body weight throughout lifespan. Different lifestyle factors are dependent on each other, e.g., physical activity, smoking, and sociodemographic factors seem to have great impact on dietary habits [6].

Several diets and surgical methods are introduced to reduce weight and maintain the weight reduction. To increase the compliance in dietary interventions, easy advices to the population are important. Not only the content of food intake, but also meal frequencies and dietary patterns have been discussed to be important factors that influence the weight development [5]. Obesity has been shown to be associated with omitted breakfasts and/or lunches and intake of dinner late in the evening [4]. In line with this, regular breakfast intake has been shown to be associated with lower risk of obesity and metabolic diseases [2,7]. In younger individuals, an invert association between meal frequency and BMI has been found [8]. When adjusted for total energy intake, lifestyle habits, and dietary factors; a meal intake of three or fewer times per day was associated with higher risk of obesity [9].

In a recently written review, regular eating habits were suggested to be of importance for weight balance, and consumption of high energy intake at the end of the day was unfavorable for weight control [10]. However, the review concluded that more studies of meal patterns and obesity are needed. Most studies are performed in the young and adolescent population, with studies on BMI and not body constitution [5,8,11]. A new syndrome called normal-weight obesity (NWO), represents subjects with a normal BMI of <25 kg/m² but with excess body fat [12,13]. This is interesting and may reflect effects of physical inactivity [13].

Our hypothesis was that low physical activity and irregular meals were associated with higher BMI, fat percentage, waist/hip ratio and a higher proportion of NWO in a middle-aged or elder population. To address this, the Swedish EpiHealth cohort of 17,724 subjects between 45-75 years was examined. The aim of the present cross-sectional population-based study was to explore the associations between physical activity and regularity of meal intakes and BMI, fat percentage, waist/hip ratio, and NWO in men and women in a middle-aged and elder Swedish population.

Subject recruitment

EpiHealth is collaboration between Lund University and Uppsala University aiming to build a national resource of a multicenter longitudinal cohort enrolling 300,000 individuals derived from the Swedish population. The EpiHealth study includes three parts: a webbased baseline questionnaire; physical tests and biological sampling at a test center; and follow-up through official Swedish registers regarding future diseases. A complete description of the study design is published [14].

All people between 45 and 75 years old who live close to the test centers (Uppsala or Malmö) and have Swedish personal security number are intended to be invited to the study by a letter containing information, activation code for the online survey, and an informed consent form. No financial compensation is offered for participants. The only exclusion criteria are unwillingness to participate or absence of personal security number. At the time point of data extraction, about 50,000 inhabitants had been invited to participate in the study. The response rate of EpiHealth is 19%. The present study participants were included 2011-2014.

Physical tests

At arrival to the EpiHealth Test Center, the participants were welcomed by a receptionist and provided oral and verbal information about the study, before the consent was signed. The participants were asked by a staff whether they had had a current infection/inflammation within the last 2 weeks and whether they had eaten during the day. If the questions were answered by “no”, weight was recorded in light clothes and with empty bladder at a scale that uses bioimpedance to calculate fat mass (Tanita, Tokyo, Japan). Height was measured. Waist circumference was measured at the umbilical level. Hip circumference was measured at the widest portion of the buttocks. Waist/hip ratio was calculated [14].

Data categorization

A database was created from the answers received in the EpiHealth questionnaire, which included questions about sociodemographic factors, family history, lifestyle habits, medical health, pharmacological treatment, as well as subjective suffering from pain and discomfort. The questionnaires used in EpiHealth are not validated questionnaires, but are similar to questionnaires used in other large population-based screening projects in Sweden (i.e., LifeGene, SCAPIS, BIG-3).

The distribution of age, BMI, fat percentage, and waist/hip ratio were not normally distributed. Age groups were sorted into 45-49 years, 50-59 years, 60-69 years, and 70-75 years. BMI, fat percentage, and waist/hip ratio were divided into two groups by the median value. Thus, BMI was grouped into ≤ 25 kg/m² and >25.0 kg/m²; fat percentage into ≤ 30% and >30%; and waist/hip ratio into ≤ 0.90 and >0.90. BMI was also divided into normal-weighted (<25 kg/m²), overweighted (25.0– 29.9 kg/m²), obese class 1 (30.0–34.9 kg/m²), and obese class 2 (≥ 35.0 kg/m²) according to the classification of the World Health Organizaton (WHO) [15]. In addition, the participants were divided into lean with a BMI of <25 kg/m² and a fat percentage in men of <20% and in women of <30%; normal-weight obesity (NWO) with a BMI<25 kg/m² and a fat percentage in men of ≥ 20% and in women of ≥ 30%; and overweight with a BMI of ≥ 25 kg/m² [12,13]. Educational level was divided into primary school, secondary school, and higher education. Occupation was divided into working, sick leave, retired, and others, including unemployed and students. Marital status was divided into single/living alone, married/cohabitation, and divorced/widowed.

Smoking habits were categorized into three groups: never smoked (or less than 100 cigarettes in total), former smokers, and current smokers. Irregular smoking was counted as current smoking. The frequency of alcohol drinking was categorized into: never, ≤ 1 time per month, 2-3 times per month, once per week, 2-3 times per week, or ≥ 4 times per week. The amount of standard drinks such a typical day was divided into 1-2 glasses, 3-4 glasses, or ≥ 5 glasses.

Lifestyle variables concerning leisure time physical activity and dietary habits were chosen to be studied for their influence on the development of body weight and body composition. Physical activity during their leisure time was estimated by cases in the questionnaire ranging from mostly sitting, light activity, walking 30 min/day, activity 30-60 min/day to strenuous activity 60 min/day [16]. Physical activity during their occupational time was not included in the present study, since only 56% of the participants were still working, and high activity during occupation time correlated with high activity during leisure time (data not shown). The frequency of food intake was estimated for breakfast, lunch, and dinner. A meal taken every day or several times a week (≥ 5 times) was considered as a “regular meal” [17].

Statistical analyses

The data was analyzed using the software SPSS, version 23.0 for Windows. Values are presented as percentages or medians (interquartile ranges (IQR)). Correlations between BMI, fat percentage, and waist/hip ratio within each subject were performed on continuous variables by Spearman´s correlation test.

Lifestyle habits intended to study (independent variables) for influence on BMI, fat percentage, waist/hip ratio, and NWO, namely, leisure time physical activity, regular intakes of breakfast, lunch, and dinner were initially examined using an unconditional logistic regression to calculate odds ratios (OR) with 95% confidence interval (CI). The reference was set to the lowest category of each variable. Calculations were thereafter adjusted for all factors in addition to sex, age, education, occupation, marital status, smoking habits, alcohol drinking frequency and amount of alcohol drinking/occasion. In addition, an interaction analysis was performed between sex and each factor in the adjusted model by including an interaction term. Logistic calculations were performed separately in women and men when statistically significant sex interactions (i) were present. The adjusted analyses were performed as a complete case-analysis. A p-value < 0.05 was considered statistically significant.

Basal characteristics

At the time of the present study, 17,724 subjects (9,936 women, 56.1%) were included in the EpiHealth study. The median age was 61 (53–67) years. Higher education was seen in 45% of the participants, and 56% of participants were still working. Seventy-three percentage of the participants were married or cohabitated. Forty-eight percentage of the subjects had never smoked, and the rest of the subjects were mostly former smokers. Less than 10% of the individuals were still smokers. The most common alcohol drinking habit was an alcohol consumption of 1-2 glasses at each occasion, with a frequency of 2–3 times a week (Table 1). The most common degree of physical activity during leisure time corresponded to 30 min of walking each day. Regular dietary habits were seen in 90%–95% of individuals, where lunch intake was the most often irregular meal intake (Table 1).

Table 1: Frequency of sociodemographic factors and lifestyle habits in the EpiHealth cohort.

Body composition

The BMI values in the whole cohort was 25 (23-28) kg/m² measured in 17,625 valid subjects (missing values=99) (Table 2). The distribution of BMI showed that 40% of the participants were categorized as normal-weighted, 43% were categorized as over weighted, 13% were categorized as obese class 1, and 3% were categorized as obese class 2. BMI correlated with fat percentage (rs=0.461, p<0.001) and waist/hip ratio (rs=0.502, p<0.001).

Table 2: Association between physical activity and regularity of meal intakes and BMI in the EpiHealth cohort.

The body fat percentage was 30 (25-36) % in 17,223 valid subjects (missing values=501) (Table 3). Body fat percentage was inversely correlated with waist/hip ratio (rs=-0.08, p<0.001). In men, 8.6% had a fat percentage of <20%, and amongst women, 46.1% had a fat percentage of <30%.

Table 3: Associations between physical activity and regularity of meal intakes and fat percentage in the EpiHealth cohort.

The waist/hip ratio was 0.90 (0.84–0.95) in 17,630 valid subjects (missing values=94) (Table 4). When dividing the cohort into three groups: lean, NWO, and overweight, the majority of participants were categorized as overweight (n=10,537) (missing values=251) (Table 5).

Table 4: Association between between physical activity and regularity of meal intakes and waist/hip ratio in the EpiHealth cohort.

Table 5: Frequency of sociodemographic factors and lifestyle habits in the EpiHealth cohort in relation to lean, normal-weight obesity and overweight.

Table 6: Association between physical activity and regularity of meal intakes and normal-weight obesity (NWO) in the EpiHealth cohort.

Physical activity

When using a logistic regression model, all forms of leisure time physical activity were inversely associated with BMI and waist/hip ratio compared to mostly sitting, whereas 30 min of walking each day and more were inversely associated with fat percentage (p for trend<0.001 for all) (Tables 2-4). When comparing lean versus NWO, higher physical activity was inversely associated with NWO, and when comparing NWO versus overweight, higher physical activity was inversely associated with overweight (p for trend<0.001 for both) (Table 6). Although a high degree of physical activity was inversely associated with higher BMI and overweight in both sexes, this effect was most pronounced in women (Tables 7 and 8).

Table 7: Interactions between sex and physical activity on BMI in the EpiHealth cohort.

Table 8: Interactions between sex and physical activity in normal-weight obesity versus overweight in the EpiHealth cohort.

Irregular meals

Irregular lunch intake was associated with higher fat percentage (OR: 1.225; 95% CI: 1.024–1.466, p=0.027) and waist/hip ratio (OR: 1.211; 95% CI: 1.033-1.421, p=0.019), whereas irregular breakfast or dinner intakes did not show any associations (Tables 3 and 4). No interactions between sex and physical activity or dietary habits could be found on fat percentage or waist/ hip ratio (data not shown). BMI or NWO did not show any associations with irregular meals (Tables 2 and 6).

The main findings of the present study was that a higher degree of leisure time physical activity was inversely associated with BMI, fat percentage, and waist/hip ratio in both sexes, although the association between physical activity and BMI was strongest in women. Irregular lunch intake showed a weak association with higher fat percentage and waist/hip ratio.

No published interventional studies could be found which estimated the effect of physical activity on body weight and body composition in elder subjects. Nevertheless, interventions with increased physical activity in adolescents have shown modest effects on reductions of BMI. However, the control groups without intervention continued to gain weight [11]. Thus, although a modest weight reduction could be found in the intervention groups [11], the benefit of physical activity is obvious compared to physical inactivity in adolescents, in line with the associations found in elder in the present study. These positive effects on weight and body composition may explain how physical activity is important to reduce morbidity and mortality in middle-aged and elder subjects [18].

Several variables may be confounders when studying the effect of dietary habits and body composition. For example, physical activity, smoking, and sociodemographic factors seem to have great impact on dietary habits [6]. A low daily meal frequency was associated with smoking, higher alcohol consumption, and lower physical activity, whereas high daily meal frequency was associated with an overall healthy lifestyle in both sexes [9].

In young adults, an additional eating occasion was associated with lower BMI and waist circumference in men, which was not found in women [19]. In middle-aged women, energy intake was increased as the meal frequency was increased [20]. Accordingly, the association between higher meal frequency and lower BMI and waist circumference is strongest in men [9,19], whereas others have found that BMI is rather associated with energy intake than meal frequency in women [20]. When the same amount of energy was eaten in three meals instead of two meals, the satiety was increased over 24 h, together with an improved oxidation of both carbohydrates and fat [21]. The intake in the morning is particularly satiating and can reduce the total energy intake for the day, whereas intake late in the evening and night lack this satiating effect and can result in greater overall daily intake [22]. Low energy intake during any meal of the day can reduce overall energy intake [22]. In the present study, we did not measure energy intake or meal frequency, just meal irregularity. Although the subjects of EpiHealth did not eat lunch, they may have compensated this by more meals during the evening/night.

In the present study of middle-aged and elder subjects, lunch intake was the most common irregular meal, and this was weakly associated with increased fat percentage and waist/hip ratio, when adjusted for confounders. In contrast, other studies have found associations between irregular meals and weight [4,5]. One explanation to the discrepancy may be that participants in the present study had overall very healthy dietary habits; only 5%–10% had irregular meal intakes. Although a great proportion of the subjects had a physical activity corresponding to 30 min of walking a day of about 40%, this is not a high activity when considered that this is the only form of physical activity during the day. These findings suggest that irregular meals are not the main cause of overweight. It is rather the low physical activity which explains that the majority of participants were over weighted or obese.

Dietary changes have great impact on circulating endocrine levels [23]. Irregular meal intake may hypothetically affect several hormones, e.g. cortisol, insulin, ghrelin, and leptin, which has been speculated to explain fluctuations in plasma glucose and insulin concentrations and lower cholesterol levels, with effects on fat percentage and fat distribution rather than weight [23,24]. Frequent or daily breakfast intake was associated with a decreased risk to develop obesity and metabolic syndrome over 18-year follow-up, compared with those who never or seldom ate breakfast [2]. In another prospective long-term study, irregularity of energy intakes over the 17-year of follow-up was associated with development of the metabolic syndrome [3].

Overweight showed similar associations versus NWO as higher BMI versus lower BMI. The previous sex differences reported, with an association between NWO and physical inactivity in men compared to lean and over weighted subjects [13], was not found in the EpiHealth cohort, where an inverse association was found between physical activity and overweight in women. The differences between the cohorts may depend on, e.g., higher age in our cohort, culture differences between the countries, and different sizes of the cohorts. In the present cohort, dividing the cohort into lean, NWO, and overweight did not lead to any further information than dividing the cohort into different BMI groups.

The strength of the present study is the large cohort of middleaged and elder subjects, and the study of fat percentage and waist/hip ratio in addition to BMI. Previous studies about the influence on meal regularity have focused on body weight, and not considered the body composition [4-6]. One limitation of the present study is the lack of information and adjustment of menopausal status. Body fat mass is increased after menopause [25], which may have affected some of the calculations. Another limitation is the absence of information about energy intake and the total number of meal intakes. The fat percentage was measured by bioimpedance. A previous study has found that this method is less reliable in elder patients, with an underestimation of fat mass [26]. That publication was published after the start of EpiHealth, why this could not be considered in the present study design. However, the same method is used in all subjects, and data are used for comparison between higher and lower fat percentage, not to use as reference values. The low inclusion prevalence is also a limitation for all parameters studied. No non-response analyses have been performed in EpiHealth, to examine the reasons for low inclusion rates, since the study inclusion has continued.

Low leisure time physical activity is associated with higher body weight, fat percentage and waist/hip ratio. Thus, physical activity has influence on both BMI and body composition in a middle-aged and elder population. Irregular lunch intake shows a weak association with higher fat percentage and waist/hip ratio, but not with body weight.

The authors want to acknowledge the steering group of EpiHealth for the data and authority to perform the study.

The data that support the findings of this study are available from the steering group of EpiHealth but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission from the steering group of EpiHealth.

BO and JM together planned and designed the study. BO performed the statistical calculations and wrote the initial draft of the manuscript. JM revised the manuscript and both authors approved the final version of the manuscript.