The obesity epidemic is one of the most important challenges for public health policy. The prevalence of excessive weight strongly affects cardiovascular health(Reference Rashid, Fuentes and Touchon1, Reference Poirier, Giles and Bray2) and contributes to a tremendous economic burden of public health(Reference Müller-Riemenschneider, Reinhold and Berghöfer3). An increasing trend of obesity has been observed during the last decade not only in the USA but also in Europe, and particularly in Spain(Reference García-Alvarez, Serra-Majem and Ribas-Barba4). Between 1995 and 2000, BMI increased by about 1 kg/m2 in men and women at the population level in north-east Spain(Reference Schröder, Elosua and Vila5).
The increase in excessive weight is accompanied by concomitant lifestyle changes, such as in diet and physical activity(Reference Nooyens, Visscher and Schuit6). Total energy intake has increased during the past 20 years in the USA, a trend driven mainly by an increase in carbohydrate consumption(Reference Briefel and Johnson7). The daily energy contribution from soft drinks increased from 1000 kJ (239 kcal) to 1230 kJ (294 kcal) over the past decade among adult North American soft drinks consumers(Reference Nielsen and Popkin8, Reference Bleich, Wang and Wang9). In Spain, daily soft drinks consumption has also increased, albeit slightly, from 44·8 ml (79 kJ/18·8 kcal) in 1991 to 72·4 ml (127 kJ/30·4 kcal) in 2001(10).
Although still inconclusive, the main body of evidence associating soft drinks consumption with BMI comes from American studies focusing on children and adolescents(Reference Malik, Schulze and Hu11–Reference Olsen and Heitmann14). Our knowledge of the impact of soft drinks consumption on BMI in the adult population is limited(Reference Chen, Appel and Loria15), particularly where the consumption is low, as in Mediterranean populations. Lifestyle differs by population and culture, and therefore the underlying dietary pattern and leisure-time physical activity involving soft drinks might vary and might have different impacts on the association of soft drinks consumption with BMI. Indeed, one might speculate that soft drinks consumption might not produce detrimental health outcomes within the context of a high-quality diet such as the Mediterranean diet. Furthermore, energy misreporting is a common problem in nutritional studies and, if not controlled for, results in biased findings about the impact of diet on BMI(Reference Heitmann and Lissner16, Reference Lissner17).
The aim of the present study was to analyse the association of soft drinks consumption with BMI and the risk of obesity, controlling for lifestyle and low energy reporting, in a representative Mediterranean population.
Materials and methods
Study participants
Data were obtained from population-based cross-sectional surveys conducted in Girona (Spain) in 2000 and 2005. These surveys of randomly selected, free-living men and women included 3058 persons aged 25–74 years in 2000 and 6352 persons aged 35–80 years in 2005. Response rates for the two surveys were 71·0 % and 71·5 %, respectively. All survey participants aged 35 to 74 years (n 8195) were included in the present study. After excluding persons with extremely low reported energy intake (<3347 kJ/<800 kcal) and extremely high value for BMI (>60 kg/m2) or reported energy consumption (corresponding to a physical activity level >2·4, which exceeds the established upper limit(Reference Black, Coward and Cole18) for strenuous work or highly active leisure behaviours), 3593 men and 3508 women were included for analysis.
The protocol was approved by the local ethics committee (CEIC-IMAS, Barcelona, Spain) and all participants were sent the results of their examination.
Anthropometric measurements
A precision scale was used for weight measurement, with subjects in underwear. Readings were rounded to the nearest 200 g. Height was measured in the standing position and measurements rounded to the nearest 5 mm. BMI was determined as weight divided by height squared (kg/m2).
Dietary assessment
Standard structured questionnaires administered by trained personnel were used to obtain information on demographic and socio-economic variables, medical history, diet and lifestyle factors, including tobacco smoking and alcohol consumption. Food consumption and nutrient intake were measured by a 168-item validated FFQ(Reference Schröder, Covas and Marrugat19) administered by a trained interviewer. The optical readable FFQ asked for usual intake over the past year of specific foods and alcoholic and non-alcoholic beverages. Participants were asked to describe their average frequency of consuming each item, using ten categories ranging from ‘almost never’ to ‘6 times/day’. Instead of standard questions on portion size based on weight or volume, the FFQ used specific medium servings, defined by natural (e.g. one orange, one slice of bread) or household units (e.g. one spoon, one cup, one glass). The FFQ included one item on soft drinks consumption (‘1 can of sugar-sweetened carbonated soda: Coca Cola, Fanta or similar, but not light’). Therefore, in the current study the term ‘soft drinks consumption’ is limited to sugar-sweetened carbonated beverages.
Measurement of diet quality
Overall diet quality was measured as adherence to the Mediterranean dietary pattern, using the Mediterranean diet score (MDS). Higher scores indicate higher adherence to the Mediterranean diet. Distribution values were calculated for all dietary components of the FFQ. The resulting MDS ranged from 10 to 30.
This operative variable for the analysis of associations between diet quality and health outcomes is calculated according to the tertile distribution of food consumption, with the exception of red wine. For cereals, fruits, vegetables, legumes, fish, olive oil and nuts, the lowest tertile is coded as 1, medium as 2 and the highest as 3. The score is inverted for meat and dairy products, with the highest tertile coded as 1 and the lowest as 3. Moderate red wine consumption (up to 20 g) is included as a favourable component in the MDS, with a score of 3. Exceeding this upper limit or reporting no red wine consumption was coded as 0.
BMR was calculated using the predictive equations based on sex, age and body weight recommended by FAO/WHO/United Nations University(Reference Schofield20). If the quotient of reported energy intake divided by the predicted BMR was <1·2, this was considered as energy under-reporting.
Other measurements
Leisure-time physical activity was measured by the Minnesota Leisure-Time Physical Activity Questionnaire and administered by a trained interviewer. This questionnaire has been previously validated for Spanish men and women(Reference Elosua, Garcia and Aguilar21, Reference Elosua, Marrugat and Molina22). Sedentary lifestyle was defined as leisure-time physical activity of less than 30 min/d.
Information on smoking habits was obtained by structured interview. Participants were categorized as non-smokers or current smokers.
Maximum education level attained was elicited and recorded for analysis as primary school, secondary school and post-secondary school.
Energy density was defined as the amount of energy (kJ) in a given weight of food (g).
Statistical analysis
Differences in continuous variables were compared using the Student t test. The χ 2 test was used for categorical variables. General linear modelling procedures (PROC GLM) in the SAS statistical software package version 9·1 (SAS Institute Inc., Cary, NC, USA) were used to estimate lifestyle, anthropometric and socio-economic variables according to categories of soft drinks consumption. For continuous variables, polynomial contrast was calculated to determine P for linear trend.
Multiple linear regression models were fitted (PROC REG procedure in SAS version 9·1) to determine the confounder-controlled association of soft drinks consumption and BMI, and produced a normal distribution of data. Multiple linear regression analysis, stratified by energy misreporting, was performed to determine the impact of energy under-reporting on the association between soft drinks consumption and BMI.
Multiple logistic regression analysis (PROC LOGISTIC procedure in SAS version 9·1) was used to assess the relationship of daily soft drinks consumption (200 ml/d) and diet quality (tertile distribution of the MDS). Differences were considered significant if P < 0·05.
Results
Less than half of the population (41·7 %) consumed soft drinks. Mean daily consumption of soft drinks consumers was 86·2 ml; monthly, weekly and daily frequency of soft drinks consumption was 55·5 % (19·0 ml/d), 27·7 % (77·9 ml/d) and 16·8 % (326·4 ml/d), respectively.
The association of soft drinks consumption with BMI, obesity and lifestyle was quite similar for both sexes. For this reason we present non-stratified results, adjusted for sex as appropriate.
Soft drinks consumers were younger, more highly educated and less prone to under-report total energy intake. They reported higher energy intakes, spent less time in leisure-time physical activity, smoked more and had a lower BMI than non-consumers (Table 1). Age decreased across frequencies of soft drinks consumption (Table 2). After controlling for sex and age, higher frequency of soft drinks consumption was associated with higher BMI and higher prevalence of sedentary lifestyle and obesity (Table 2).
Table 1 General characteristics of the study population according to frequency of soft drinks consumptionFootnote *: random sample of the 35- to 74-year-old population (3910 men and 4285 women), Girona, Spain
LTPA, leisure-time physical activity; MET, metabolic equivalent task.
* Results are expressed as percentage of subjects and 95 % confidence interval or mean and standard deviation. Significance of P between soft drinks consumers and soft drinks non-consumers was determined by the Student t test (continuous variables) or logistical regression analysis (categorical variables).
† Energy intake:BMR <1·2.
‡ Energy density was calculated as energy intake from all foods consumed (kcal) divided by weight of foods consumed (g); 1 kcal = 4·184 kJ.
§ More than primary school.
Table 2 General characteristics of the population by frequency of soft drinks consumptionFootnote *: random sample of the 35- to 74-year-old population (3910 men and 4285 women), Girona, Spain
LTPA, leisure-time physical activity; MET, metabolic equivalent task.
* Results are expressed as percentage of subjects or mean and 95 % confidence interval. Sex- and age-adjusted ANOVA was used to estimate variables according to frequency of soft drinks consumption.
† BMI ≥ 30 kg/m2.
‡ BMI = 25·0–29·9 kg/m2.
§ Less than 30 min LTPA/d.
|| More than primary school.
Soft drinks consumption was directly associated with energy intake, energy density and intakes of carbohydrates, pastry/sweets and high-fat dairy products (Table 3). A decrease occurred across categories of soft drinks consumption in the ratio of unsaturated to saturated fat and intakes of protein, total fat, fibre, olive oil, low-fat dairy products, fish, vegetables, fruits, nuts, poultry/rabbit and legumes (Table 3).
Table 3 Age- and sex-adjusted daily food and nutrient intakes by category of soft drinks consumptionFootnote *: random sample of the 35- to 74-year-old population (3910 men and 4285 women), Girona, Spain
* Results are expressed as mean or percentage of subjects and 95 % confidence interval. Sex- and age-adjusted ANOVA was used to estimate variables according to frequency of soft drinks consumption.
† Energy density was calculated as energy intake from all foods consumed (kcal) divided by weight of foods consumed (g); 1 kcal = 4·184 kJ.
‡ Energy intake:BMR <1·2.
§ Percentage of total energy intake.
Multiple linear regression analysis – adjusted for sex, age, educational status, leisure-time physical activity, energy intake, smoking, alcohol consumption and energy under-reporting – revealed that a 100 ml increment in soft drinks consumption was associated with an increase of 0·213 kg/m2 in BMI (P < 0·001; Table 4). This association was unchanged in both magnitude and direction in plausible energy reporters but was attenuated in low energy reporters. Multivariate ANOVA adjusted for sex and age revealed a positive association of soft drinks consumption across BMI categories (normal weight, overweight and obese; P < 0·001). Participants with normal weight consumed 31·2 ml/d whereas their obese peers drank 42·4 ml/d (P = 0·003). Additionally controlling for energy intake, leisure-time physical activity, smoking, educational level, diet quality and energy under-reporting attenuated this association slightly (P for linear trend = 0·007; soft drinks consumption by normal weight v. obese participants: 32·2 ml/d v. 41·5 ml/d; P = 0·021). Figure 1 shows the association of soft drinks consumption with overall diet quality, defined by MDS. Adherence to a healthy diet decreased with daily soft drinks consumption (P < 0·001).
Table 4 Regression coefficient and 95 % confidence interval of the association between soft drinks consumption and BMIFootnote * among a random sample of the 35- to 74-year-old population (3910 men and 4285 women), Girona, Spain
* Multiple linear regression models were fitted to analyse the association of soft drinks consumption and BMI.
† Adjusted for sex, age, leisure-time physical activity, educational level, smoking, alcohol consumption, diet quality (Mediterranean diet score), energy (energy from soft drinks excluded) and energy under-reporting (energy intake:BMR <1·2).
‡ 100 ml corresponds to 0·5 drinks/d.
§ Adjusted for sex, age, leisure-time physical activity, educational level, smoking, alcohol consumption, diet quality (Mediterranean diet score) and energy (energy from soft drinks excluded).
Fig. 1 Odds ratio and 95 % confidence interval of daily soft drinks consumption and adherence to the Mediterranean diet (tertile distribution of the Mediterranean diet score) among a random sample of the 35- to 74-year-old population (3910 men and 4285 women), Girona, Spain. Odds ratios were adjusted for sex, age, leisure-time physical activity, educational level, smoking, alcohol consumption, diet quality (Mediterranean diet score), energy (energy from soft drinks excluded) and energy under-reporting (energy intake:BMR <1·2). P for linear trend = 0·001
Discussion
Soft drinks consumption was directly associated with BMI. Stratifying by implausibly low and plausible energy reporters revealed a significant association of soft drinks consumption and BMI in plausible reporters but not in under-reporters. Furthermore, daily soft drinks consumption was associated with low diet quality.
Soft drinks consumption in the USA constitutes 7·1 % of total energy intake(Reference Guthrie and Morton23). In contrast, in the present population soft drinks consumption can be considered low, since it contributes only 61 kJ (14·5 kcal) to total energy intake (0·6 % of the total energy supply). Furthermore, the average daily amount (86·8 ml) reported by soft drinks consumers falls within the range of the recently proposed tolerable intake of soft drinks(Reference Popkin, Armstrong and Bray24). The Mediterranean dietary pattern is associated with favourable health outcomes(Reference Giugliano and Esposito25–Reference Martinez-Gonzalez, Bes-Rastrollo and Serra-Majem27). It is conceivable that soft drinks consumption at these amounts is unlikely to replace healthy food choices in this context. Furthermore, moderate soft drinks consumption as part of a healthy lifestyle should not affect weight gain, cardiovascular health or BMI. However, our data showed that frequent soft drinks consumption increased the risk of low adherence to the Mediterranean diet and was associated with a cluster of unhealthy food preferences. The observation that unhealthy food choices are associated with soft drinks consumption is in line with previously published reports from younger populations(Reference Ballew, Kuester and Gillespie28, Reference Yamada, Murakami and Sasaki29).
In the present population there was a positive association, controlled for potential confounders, between soft drinks consumption and BMI. Limited evidence from other cross-sectional and prospective epidemiological studies in adult populations, particularly from the USA, has indicated a modest positive association(Reference Malik, Schulze and Hu11–Reference Chen, Appel and Loria15). Chen et al. reported a significant weight loss associated with a reduction of 1 serving of soft drinks per day after 18 months(Reference Chen, Appel and Loria15). A recently published Spanish study of male and female university alumni found a significant positive association between soft drinks consumption and weight gain among adults who had gained more than 3 kg weight in the 5 years before the study(Reference Bes-Rastrollo, Sánchez-Villegas and Gómez-Gracia30). In the present study, a 100 ml increment in soft drinks consumption was associated with an increase of 0·21 kg/m2 in BMI. Although the magnitude of the association is modest, the fact that it derives from a single food item, rather than a complex dietary pattern, reflects a considerable impact on BMI in this population. The association of BMI and soft drinks consumption was only slightly attenuated after controlling for overall diet quality. This finding indicates that soft drinks consumption affects BMI independently of the magnitude of diet quality as measured by adherence to the Mediterranean diet. Even within the context of an overall healthy diet, soft drinks consumption adversely affects BMI.
Several mechanisms linking soft drinks consumption and weight gain have been proposed(Reference Bawa31). The predominant evidence points to an imbalance in energy intake, as soft drinks mediate less satiation and low satiety(Reference Bachman, Baranowski and Nicklas32). In the present population, higher soft drinks consumption was not associated with energy compensation in the overall diet. By definition, excess energy intake through soft drinks consumption without compensatory energy expenditure is a formula for weight gain in the long run.
Unfortunately, energy under-reporting is a common problem for the analysis of associations between diet and health outcomes, particularly BMI(Reference Heitmann and Lissner16, Reference Lissner17). It has been shown that obese subjects tend to selectively under- or overestimate selected foods(Reference Heitmann and Lissner16, Reference Lissner17). In the present study, prevalence of low energy reporting decreased with frequent soft drinks consumption. Thus, it was not surprising that soft drinks consumption was positively associated with BMI in plausible reporters but exhibited a null association in those who reported low energy consumption. Our results indicate that controlling for energy under-reporting is essential to avoid bias in the association of foods or diet and BMI.
A limitation of the present study’s cross-sectional design is that causality cannot be drawn between the variables studied. Misreporting is an acknowledged source of measurement error in prospective or retrospective methods of dietary assessment using self-reported food intake records. Furthermore, total consumption of sugar-sweetened drinks is underestimated in this population because the term ‘soft drink’ did not include non-carbonated sugar-sweetened beverages. The strengths of the present study include the relatively large sample size, the population-based design and the consideration of potential confounders in the analysis, including physical activity, smoking, socio-economic status and energy under-reporting.
In conclusion, frequent consumers of soft drinks tended towards a more sedentary lifestyle and a cluster of unhealthy dietary habits. The risk of low adherence to the Mediterranean diet increased significantly with daily soft drinks consumption. Higher BMI was found in more frequent consumers of soft drinks as compared with their non-consuming peers, after controlling for several potential confounders. Low energy reporters showed a null association between soft drinks consumption and BMI, indicating that energy misreporting is a strong potential confounder for analysis of the relationship between soft drinks consumption and BMI.
Acknowledgements
This work was supported by a grant from the Instituto de Salud Carlos III FEDER (PI080439) and a joint contract of the Instituto de Salud Carlos III and the Health Department of the Catalan Government (Generalitat de Catalunya), CP 03/00115. The CIBEROBN is an initiative from the Instituto de Salud Carlos III, Madrid, Spain. The authors declare no conflict of interest. E.B. and M.D.-N. contributed equally to the study, and each can be regarded as first author of the present paper. The manuscript was prepared by E.B. and M.D.-N., with significant input and feedback from all co-authors. H.S. contributed to the discussion and interpretation of results and writing of the manuscript. R.P.-L., T.R.-G., A.R., G.G.-Z., J.M.-O., S.V.-H. and M.P.-C. provided expertise in data analysis and in interpretation and discussion of the results. They made substantial suggestions on the manuscript. The authors appreciate the English revision by Elaine M. Lilly, PhD (Writers First Aid).
