High levels of LDL-cholesterol (LDL-C) in childhood and the onset of atherosclerosis in early life( Reference Saland 1 ) could result in adult dyslipidaemias( Reference Nicklas, von Duvillard and Berenson 2 ), which are important cardiovascular risk factors. Therefore, evidence for factors influencing lipid profiles is necessary for public health protection and promotion.
Dietary modifications that lower atherogenic lipids and lipoproteins are effective in the prevention and treatment of CVD risk( Reference Flock, Green and Kris-Etherton 3 ). Nonetheless, the optimal dietary pattern(s) to restrain atherosclerosis progression remains to be identified( Reference Mozaffarian, Rimm and Herrington 4 ). For instance, a high intake of total carbohydrates has been suggested to lower HDL-cholesterol (HDL-C) concentration and to increase TAG concentration in adults( Reference Ma, Li and Chiriboga 5 ), and a protein-rich diet low in saturated fat has been shown to significantly decrease the concentrations of LDL-C, TAG and total cholesterol (TC) in comparison to a carbohydrate-rich diet and a diet rich in unsaturated fat( Reference Appel, Sacks and Carey 6 ). However, findings about the role of dietary fat, mainly that of saturated fat, in CVD risk are still controversial( Reference Siri-Tarino, Sun and Hu 7 ) because of individual variability in serum lipid response to changes in dietary saturated fat and cholesterol( Reference Flock, Green and Kris-Etherton 3 ). Additionally, obesity has been reported to strongly affect serum lipid response to diet( Reference Flock, Green and Kris-Etherton 3 ), making obese individuals less responsive to dietary interventions aimed to improve serum lipid profile.
Adolescence is a key period in life because of the growth spurt and sexual maturation that take place; therefore, having a healthy diet is essential to achieve optimal development and to prevent the appearance of chronic diseases later in life( Reference Bertheke Post, de Vente and Kemper 8 ). Furthermore, associations between macronutrient intake and serum lipids have mainly been investigated among adults, and there is a lack of the literature addressing this topic in adolescents. As these associations have not been examined yet among adolescents, we hypothesised that macronutrient intake was associated with serum lipids in a sample of healthy European adolescents and that body adiposity may exert a key role in this association. Therefore, the aims of the present study were (1) to investigate the relationships between macronutrient intake and serum lipid profile in European adolescents and (2) to assess the role of body fat-related variables in these associations.
Materials and methods
The present study sample was derived from the cross-sectional multi-centre HELENA (Healthy Lifestyle in Europe by Nutrition in Adolescence) study (n 3528) carried out in adolescents (12·5–17·5 years) between 2006 and 2007 in ten European cities (Athens and Heraklion in Greece, Dortmund in Germany, Ghent in Belgium, Lille in France, Pécs in Hungary, Rome in Italy, Stockholm in Sweden, Vienna in Austria and Zaragoza in Spain). General HELENA procedures, characteristics and inclusion criteria can be found elsewhere( Reference Moreno, De Henauw and Gonzalez-Gross 9 , Reference Moreno, Gonzalez-Gross and Kersting 10 ). The present study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving human subjects were approved by the local ethical committee at each study centre( Reference Beghin, Castera and Manios 11 ). Written informed consent was obtained from all adolescents and their parents.
Participants with complete data on TAG, TC, HDL-C, LDL-C, apoA1, apoB and two 24 h dietary recalls (24-HDR) were included (n 454; 44 % boys). Decreases from the original sample size are partially explained by the fact that blood samples were randomly drawn only in one-third of the HELENA participants and partially from the fact that Heraklion and Pécs were excluded from the 24-HDR analyses due to logistical reasons; therefore, eight out of the ten study centres were included in the 24-HDR analyses, resulting in a sample size decrease. Excluded participants (n 3074) were significantly (P< 0·05) older, heavier and had a higher mean BMI than those included in the present study (data not shown).
Macronutrient intake
Dietary intake was assessed using a self-administered computer-based tool called HELENA-DIAT (Dietary Assessment Tool), based on the previously developed software Young Adolescents’ Nutrition Assessment on Computer (YANA-C) that was shown to be appropriate in assessing dietary information of European adolescents( Reference Vereecken, Covents and Matthys 12 , Reference Vereecken, Covents and Sichert-Hellert 13 ). The software consists of a single, structured 24-HDR according to six meal occasions. Adolescents were asked to recall all food and drinks consumed the previous day. Within a time span of 2 weeks, two non-consecutive 24-HDR were obtained from each participant during school time and assisted by fieldworkers. Therefore, no information on Fridays and Saturdays was collected.
The German Food Code and Nutrition Data Base (Bundeslebensmittelschlüssel, BLS version II.3.1)( Reference Dehne, Klemm and Henseler 14 ) was used to calculate energy and nutrient intakes. Usual food and nutrient intakes were estimated by the multiple source method in order to account for within-person variability( Reference Harttig, Haubrock and Knuppel 15 ). Energy intake was estimated in kJ/d and macronutrient intake (fat, protein and carbohydrate) in g/d. Subsequently, intakes of each macronutrient were divided by energy intake and are expressed as g/4180 kJ per d (1000 kcal per d) to account for total energy intake( Reference Willett, Howe and Kushi 16 ). Additionally, the monounsaturated:saturated fat ratio (M:S), the polyunsaturated:saturated fat ratio (P:S) and the cholesterol–saturated fat index (CSI) were computed as follows( Reference Connor, Gustafson and Artaud-Wild 17 ):
$$\begin{eqnarray} CSI = (1\cdot 01\times g\,saturated\,fat) + (0\cdot 05\times mg\,cholesterol). \end{eqnarray}$$
Diet Quality Index for Adolescents
A previously validated Diet Quality Index for Adolescents (DQI-A)( Reference Vyncke, Cruz Fernandez and Fajo-Pascual 18 ) was used to adjust for all dietary factors simultaneously. The technical aspects regarding the development of the DQI-A have been published elsewhere( Reference Vyncke, Cruz Fernandez and Fajo-Pascual 18 ).
Physical examinations
All anthropometric measurements were taken following a standardised protocol described elsewhere( Reference Nagy, Vicente-Rodriguez and Manios 19 ). Weight and height were measured in underwear and barefoot using an electronic scale (Type SECA 861) and a stadiometer (Type SECA 225). BMI was calculated as body weight (in kg) divided by the square of height (in m) and was categorised as described by Cole et al. ( Reference Cole, Bellizzi and Flegal 20 , Reference Cole, Flegal and Nicholls 21 ). Skinfold thicknesses were measured with a calliper (Holtain Ltd) in triplicate on the left side at the biceps, triceps, subscapular and suprailiac sites. Waist circumference was measured at the midpoint between the lowest rib and the iliac crest using an anthropometric tape (SECA 200). The waist:height ratio (WHeR) was calculated.
Blood sampling
Blood sampling procedures have been described elsewhere( Reference Gonzalez-Gross, Breidenassel and Gomez-Martinez 22 ). Briefly, blood samples were drawn after a 10 h overnight fast and analysed in centralised laboratories. Serum TAG, TC, HDL-C, TAG and LDL-C concentrations were measured on a Dimension RxL clinical chemistry system (Dade Behring) with enzymatic methods using the manufacturer's reagents and instructions. ApoB and apoA1 were measured in an immunochemical reaction with a BN II analyser (Dade Behring), according to the manufacturer's instructions. The technique consists of the following: the proteins contained in the serum sample form immune complexes with specific antibodies. These complexes scatter a beam of light when it passes through the sample. As the intensity of the scattered light is proportional to the concentration of the relevant protein in the sample, the result is evaluated by comparison with a standard of known concentration. Quantitative evaluation was made by comparison with standard concentrations. Intra-assay CV for all blood variables were < 3·9 %, and inter-assay CV were < 4·3 %. The TC:HDL-C and the apoB:apoA1 ratios were computed.
Education
Maternal education was used as a proxy of socio-economic status and was assessed via a questionnaire according to the following four categories: (1) lower education; (2) lower secondary education; (3) higher secondary education; (4) higher education/university degree.
Sedentary behaviours
Average time spent in two sedentary behaviours (television viewing and playing with videogames) was estimated by means of a self-administered questionnaire that has been previously found to demonstrate good reliability( Reference Rey-Lopez, Ruiz and Ortega 23 ).
Physical activity
Physical activity (PA) was objectively measured by uniaxial accelerometers during seven consecutive days (Actigraph MTI, model GT1M; Manufacturing Technology, Inc.)( Reference Ruiz, Ortega and Martinez-Gomez 24 ). At least 3 d of recording, with a minimum of 8 h registration per d, was set as an inclusion criterion. The time sampling interval was set at 15 s. The time spent in moderate-to-vigorous PA (>3 metabolic equivalents) was calculated on the basis of the following cut-off point: ≥ 2000 counts/min for moderate-to-vigorous PA( Reference Ruiz, Ortega and Martinez-Gomez 24 , Reference Ekelund, Sardinha and Anderssen 25 ).
Statistical analysis
The normality of all variables was checked and non-normally distributed variables (TAG, TC, HDL-C, LDL-C, TC:HDL-C ratio, apoB:apoA1 ratio, fat intake and P:S ratio) were log transformed before the analysis. Normality for the CSI was reached by using the power of 2. The M:S ratio was converted to 1/(M:S). Differences across the groups were tested by means of the independent-samples t test for normally distributed variables and the Mann–Whitney U test for non-normally distributed variables. The χ2 test was applied for categorical variables.
Multi-level linear regression analyses were performed to investigate the associations between the intakes of macronutrients (independent variables) and plasma lipid concentrations (dependent variables). As no interaction by sex was found, the analyses were conducted with boys and girls combined. Study centre was included as the random intercept. Sex, age, maternal education, sum of four skinfolds, moderate-to-vigorous PA, sedentary behaviours and DQI-A were entered as covariates. Collinearity tests showed no collinearity among the covariates.
Since serum lipid profile has previously been associated with body fat in the HELENA adolescents( Reference Spinneker, Egert and Gonzalez-Gross 26 ) and excess adiposity has been shown to have an influence on serum lipid response to diet( Reference Flock, Green and Kris-Etherton 3 ), participants were categorised into low and high body fat content according to three body fat indicators, i.e. z-score of BMI and sum of skinfolds as the measures of general body fatness, and the WHeR as an indicator of central adiposity. These cut-offs were calculated specifically by sex and by 1-year groups (12·5–13·49, 13·5–14·49, 14·5–15·49, 15·5–16·49 and 16·5–17·5) based on the median of each subgroup. Multi-level linear regression analyses on the associations between macronutrient intake and serum lipid concentrations were performed separately for each group of low/high body fat indicator and adjusted for potential confounders, i.e. sex, age, maternal education, sum of four skinfolds, moderate-to-vigorous PA, sedentary behaviours and DQI-A, where study centre was entered as the random intercept. No collinearity was observed among the covariates. The level of statistical significance was controlled for multiple testing (0·05/number of tests = 0·05/8 = 0·006); therefore, statistical significance was considered at P≤ 0·006. Multi-level linear regression analyses were re-run by using tertiles of protein, carbohydrate and fat intake. Bonferroni correction was used for the post hoc multiple comparison test, and statistical significance was set at P< 0·05. Statistical analysis was performed using the statistical software package STATA version 12.0 (Stata Corporation).
Results
The main characteristics of the study sample are shown in Table 1. Table 2 presents the means and medians of dietary intake and blood lipid levels according to high/low body WHeR, i.e. above or below the sex- and age-specific median-based cut-offs of the WHeR. Adolescents in the high-WHeR group showed significantly higher protein intake (g/4180 kJ) and percentage of protein intake, M:S ratio, TAG, TC:HDL-C ratio and apoB:apoA1 ratio and lower total energy intake than adolescents in the low-WHeR group.
Descriptive characteristics of the study sample stratified by sex (Mean values and standard deviations; percentages; medians and 25th–75th percentiles)
MVPA, moderate-to-vigorous physical activity; DQI-A, Diet Quality Index for Adolescents; TC, total cholesterol; HDL-C, HDL-cholesterol; LDL-C, LDL-cholesterol.
* P< 0·05 (independent-samples t test).
† BMI categories as described by Cole et al. ( Reference Cole, Bellizzi and Flegal 20 , Reference Cole, Flegal and Nicholls 21 ).
‡ P< 0·05 (χ2 test).
§ P< 0·05 (Mann–Whitney U test).
Dietary characteristics and serum lipid parameters stratified by high v. low waist:height ratio‡ (Mean values and standard deviations; medians and 25th–75th percentiles)
TC, total cholesterol; HDL-C, HDL-cholesterol; LDL-C, LDL-cholesterol.
* Mean value was significantly different from that of adolescents in the low-adiposity waist:height ratio group (P< 0·05; independent-samples t test).
† Median value was significantly different from that of adolescents in the low-adiposity waist:height ratio group (P< 0·05; Mann–Whitney U test).
‡ High- and low-waist:height ratio groups were defined by means of sex- and age-specific medians.
The associations between macronutrient intake (g/4180 kJ) and serum lipid profile are shown in Table 3. Carbohydrate intake was inversely associated with HDL-C (P= 0·001), and a trend towards significance (P= 0·010) was observed for apoA1. Inverse associations were found between fat intake and TAG (P< 0·001) and TC:HDL-C ratio (P= 0·005). LDL-C, apoB and apoB:apoA1 ratio were not significantly associated, but an inverse trend towards significance was observed. In addition, a positive trend towards significance was found between fat intake and HDL-C (P= 0·047). Protein intake was not significantly associated with TAG, but an inverse trend towards significance was observed (P= 0·027).
Multi-level regression analysis addressing the associations between intakes of protein, carbohydrates and fat, monounsaturated:saturated fat ratio, polyunsaturated:saturated fat ratio and cholesterol–saturated fat index and serum lipid profile (β-Coefficients and 95 % confidence intervals)
TC, total cholesterol; HDL-C, HDL-cholesterol; LDL-C, LDL-cholesterol.
* P< 0·006 (Bonferroni correction).
† Adjusted for sex, age, study centre, maternal education, sum of four skinfolds, moderate-to-vigorous physical activity, sedentary behaviours and dietary quality index.
Analyses were conducted for the three body fat indicators; however, the results have been focused on only one measure of adiposity, specifically central adiposity, to ease its interpretation. Table 4 presents the associations between macronutrient intake and blood lipid profile for the low and high WHeR. Carbohydrate intake was inversely associated (P= 0·001) with HDL-C in adolescents with the high WHeR and a negative trend was found for apoA1 (P= 0·043), but not in those with the low WHeR. A positive trend towards significance (P= 0·011) was observed between the intake of carbohydrate and TAG concentration only in those adolescents within the high WHeR group. An inverse trend between fat intake and TAG (P= 0·008) and TC:HDL-C ratio (P= 0·044) was found in adolescents with the high WHeR. Although not significantly associated, inverse trends between protein intake and TAG (P= 0·021) and LDL-C (P= 0·049) concentrations were observed only in adolescents with the high WHeR. The results were consistent with the other measures of adiposity, namely sum of four skinfolds and z-score of BMI (data not shown).
Multi-level regression analysis addressing the associations between macronutrient intake and serum lipid profile categorised by high v. low waist:height ratio (WHeR)‡ (β-Coefficients and 95 % confidence intervals)
TC, total cholesterol; HDL-C, HDL-cholesterol; LDL-C, LDL-cholesterol.
* P< 0·006 (Bonferroni correction).
† Adjusted by sex, age, study centre, maternal education, moderate-to-vigorous physical activity, sedentary behaviours and diet quality index for adolescents.
‡ High and low waist:height ratio was defined by means of sex- and age-specific medians.
Fig. 1 shows the means of TAG, TC, HDL-C and LDL-C concentrations by tertiles of fat intake in adolescents with the high and low WHeR. Overall, adolescents with the high WHeR in the lower tertile of fat intake (tertile 1) showed clinically adverse values of serum lipid concentrations, i.e. higher TAG, TC and LDL-C concentrations, and lower HDL-C concentration, compared with those in the higher tertile of fat intake (tertile 3). Within the upper tertile of fat intake, adolescents with the high WHeR had better values of blood parameters, whereas slight differences were observed across the tertiles of fat intake in adolescents with lower central adiposity levels. Indeed, adolescents with the high WHeR (Fig. 1(a)) in the lowest tertile of fat intake showed significantly higher TAG concentrations than those in the upper tertile of fat intake (P< 0·05). Furthermore, a significant interaction (P< 0·05) was observed between fat intake and WHeR for TAG (Fig. 1(a)). In contrast, adolescents with high adiposity levels, i.e. high WHeR, in the upper tertile of carbohydrate intake showed significantly higher TAG levels than those in the lower tertile of carbohydrate intake (P< 0·05, data not shown). No significant differences were observed across tertiles of protein intake (data not shown). Identical findings across the tertiles of fat intake were observed in adolescents with the high/low sum of skinfolds.
TAG (a), total cholesterol (b), HDL-cholesterol (HDL-C) (c) and LDL-cholesterol (LDL-C) (d) by tertiles of fat intake for high (–○–) and low (–●–) waist:height ratio (WHeR) after adjustment for covariates: age; sex; study centre; socio-economic status; moderate-to-vigorous physical activity; sedentary behaviours; diet quality index for adolescents. Values are means, with their standard errors represented by vertical bars. To convert TAG in mg/dl to mmol/l, multiply by 0·0113. To convert cholesterol in mg/dl to mmol/l, multiply by 0·0259. * Mean value was significantly different from that of adolescents in the higher tertile of fat intake (P< 0·05; Bonferroni's correction for post hoc multiple comparisons). Unlogged values are shown for easier interpretability. Median fat intake: tertile 1 – 32·1 g/4180 kJ (1000 kcal) (low WHeR), 33·6 g/4180 kJ (1000 kcal) (high WHeR); tertile 2 – 40·8 g/4180 kJ (1000 kcal) (low WHeR), 41·9 g/4180 kJ (1000 kcal) (high WHeR); tertile 3 – 50·0 g/4180 kJ (1000 kcal) (low WHeR), 51·4 g/4180 kJ (1000 kcal) (high WHeR).
Discussion
The present study examined the associations between energy-adjusted macronutrient intake and serum lipid profile, as well as the potential role that body adiposity may exert on these associations among healthy European adolescents. Overall, the results suggested that dietary fat has a beneficial role in serum lipid levels by lowering serum TAG levels and TC:HDL-C ratio, whereas carbohydrate intake was adversely associated with lipid profile by decreasing serum HDL-C concentrations. These above-mentioned associations varied according to the body fat status of adolescents, i.e. significant associations and trends towards significance between intakes of fat and carbohydrate and blood lipids were observed mainly among adolescents in the high-body fat group. To the best of our knowledge, the present study is a novel study as it is the first to address such relationships among adolescents.
The associations between carbohydrate intake and HDL-C add further evidence to the complex and adverse role that dietary carbohydrates appear to play in serum lipid profile( Reference Ma, Li and Chiriboga 5 ). Indeed, a high intake of total carbohydrate is also associated with lower HDL-C and higher TAG concentrations in adults( Reference Ma, Li and Chiriboga 5 ) and children( Reference Ruottinen, Ronnemaa and Niinikoski 27 ). Nevertheless, it is important to take into account that we did not discriminate among the types of carbohydrates, i.e. simple and complex carbohydrates, meaning that the observed associations may have differed among the subtypes of carbohydrates. However, it is known that serum lipid levels are controlled not only by dietary carbohydrates but also by dietary proteins( Reference Oda 28 ). Although available data addressing the associations between dietary protein intake and serum lipids are still limited, vegetal sources of protein per se have been shown to lower plasma cholesterol concentrations( Reference Potter 29 ). Furthermore, Appel et al. ( Reference Appel, Sacks and Carey 6 ) observed that a healthy diet, rich in protein and low in saturated fat, significantly decreased the concentrations of LDL-C, TAG and TC among adults compared with a carbohydrate-rich diet and a diet rich in unsaturated fat. The inverse association that we found between protein intake and TAG concentration is partially in concordance with these findings.
The role of fat intake in serum lipid levels differs according to the type of fat consumed. Indeed, the fatty acid profile of the diet seems to be the major determinant of serum cholesterol concentrations( Reference Lichtenstein 30 ). The findings from a follow-up study( Reference Jakobsen, O'Reilly and Heitmann 31 ) have suggested that replacing dietary saturated fat with polyunsaturated fat rather than with monounsaturated fat or carbohydrates protects middle-aged and older men and women from the risk of CHD. Other studies have observed that replacement of dietary saturated fat with polyunsaturated fat, monounsaturated fat and/or carbohydrates reduces LDL-C concentration( Reference Berglund, Lefevre and Ginsberg 32 , Reference Lichtenstein, Matthan and Jalbert 33 ). PUFA intake decreases LDL-C concentration( Reference Lichtenstein 30 ) and MUFA intake decreases the TC:HDL-C ratio when isoenergetically compared with SFA intake( Reference Gillingham, Harris-Janz and Jones 34 ). Monounsaturated fat-rich diets have been shown to have comparable effects on serum lipid concentrations with diets rich in PUFA, although they tend to reduce TAG concentration less, but elevate HDL-C concentration more than PUFA-rich diets( Reference Gillingham, Harris-Janz and Jones 34 ). Accordingly, further analysis carried out in the present study sample showed a positive association between monounsaturated fat intake and HDL-C and apoA1. For this reason, the inverse association observed between total fat intake and TAG and TC:HDL-C ratio could be partially explained by the presence of polyunsaturated and monounsaturated fat, although no significant associations were observed between the M:S and P:S ratios and serum lipids. However, adolescents with higher levels of adiposity showed significantly higher intakes of monounsaturated fat and a higher M:S ratio. The results from a recent meta-regression( Reference Schwingshackl and Hoffmann 35 ) have revealed that increases in HDL-C concentrations are associated with higher amounts of total fat mainly derived from monounsaturated fat in high-fat diets, whereas higher intakes of carbohydrates are associated with increases in TAG levels. Lyu et al. ( Reference Lyu, Yeh and Lichtenstein 36 ) also found a positive association between total fat intake and apoA1 in both men and women. Such findings could denote a positive role of dietary fat in different fractions of serum lipids and are consistent with several experimental studies that observed unfavourable effects of low-fat diets on HDL-C, TC:HDL-C ratio and postprandial TAG concentrations in women when compared with men( Reference Cobb, Greenspan and Timmons 37 – Reference Lichtenstein, Ausman and Jalbert 40 ).
Focusing on saturated fat, a meta-analysis of prospective cohort studies( Reference Siri-Tarino, Sun and Hu 7 ) did not find significant evidence to conclude that dietary saturated fat intake was associated with the increased risk of CVD. However, these findings should be interpreted with caution as, according to Kromhout et al. ( Reference Kromhout, Geleijnse and Menotti 41 ), existing sources of error in both dietary exposure and effect measure might have attenuated the correlation between dietary saturated fat intake and serum cholesterol concentration, leading to a correlation close to zero. Nevertheless, Mozaffarian et al. ( Reference Mozaffarian, Rimm and Herrington 4 ) observed that a greater intake of saturated fat was associated with higher HDL-C and apoA1 concentrations, and lower TAG concentration and TC:HDL-C ratio in postmenopausal women. It seems that not all SFA have identical effects on serum cholesterol levels( Reference Lichtenstein 30 ), suggesting that the effects of saturated fat intake on serum lipids vary according to the specific SFA consumed( Reference Mensink, Zock and Kester 42 ). The findings from a study conducted among Swedish adolescents revealed significant inverse associations between the dietary content of SFA with a chain length of four to fifteen carbon atoms and the serum concentrations of TC and apoB( Reference Samuelson, Bratteby and Mohsen 43 ). A meta-analysis of controlled feeding experiments has shown that SFA with twelve, fourteen, sixteen and eighteen carbon atoms increased the concentration of HDL-C when they isoenergetically replaced carbohydrate( Reference Micha and Mozaffarian 44 ). It should be noted that the increases in HDL-C concentration were greater as the chain length decreased. Overall, the TC:HDL-C ratio was not significantly affected by SFA with fourteen, sixteen and eighteen carbon atoms; however, it significantly decreased when SFA with twelve carbon atoms replaced carbohydrate( Reference Micha and Mozaffarian 44 ). In addition to the effect of the chain length, there may also be an effect of the source of saturated fat on TC and HDL-C concentrations. For example, despite its animal origin, milk fat elevates serum HDL-C concentration( Reference Rice, Cifelli and Pikosky 45 ). In addition, the unique position of SFA in milk fat, which is typically the position of unsaturated fatty acids in plant oils, may also affect postprandial metabolism, leading to the prevention of elevated serum TC and TAG concentrations( Reference Rice, Cifelli and Pikosky 45 ).
Variability in lipid response to diet is affected by numerous factors, but excess adiposity seems to be one of the strongest factors( Reference Flock, Green and Kris-Etherton 3 ). Our findings showed that associations between macronutrient intake and serum lipids varied by body fat status in the present study sample of healthy adolescents regardless of the definition employed, i.e. general body fatness (z-score of BMI and sum of skinfolds) or central body fat (WHeR). Previous reports have pointed out that adiposity plays a key role in these associations among adults( Reference Lyu, Yeh and Lichtenstein 36 , Reference Clifton, Abbey and Noakes 46 ). Lyu et al. ( Reference Lyu, Yeh and Lichtenstein 36 ) observed that general body fatness, but also body fat distribution, might exert different effects on HDL-C subclasses. Obese individuals have shown lower responses to dietary interventions focused on improving their serum lipid profile( Reference Flock, Green and Kris-Etherton 3 ). Indeed, Denke et al. ( Reference Denke, Adams-Huet and Nguyen 47 ) found that the response to a LDL-C-lowering diet was lower in obese participants than in those with a lower BMI ( < 21 kg/m2). However, our findings showed significant associations between macronutrient intake and blood lipids mainly in the high-adiposity groups, which initially would be in contrast to previous literature.
There were several limitations to our findings. Due to the cross-sectional nature of the study, we cannot determine causality. Hormonal changes during the menstrual cycle may influence serum lipid concentrations; however, blood samples were not taken at the consistent time of the menstrual cycle to account for that factor. Although the self-administered 24-HDR used to assess dietary intake is subject to measurement errors as occurs with other self-reporting methods, it has been shown to be appropriate to collect detailed dietary data in adolescents( Reference Vereecken, Covents and Matthys 12 , Reference Vereecken, Covents and Sichert-Hellert 13 ). Collection of dietary data for more than 2 d would have been desirable to compensate for day-to-day variability( Reference Thompson, Subar, Coulston and Boushey 48 ); nevertheless, dietary information was corrected for within-person variability to partially mitigate such limitation( Reference Dodd, Guenther and Freedman 49 ). Additionally, our sample included under-reporters; however, results did not change when they were withdrawn from the analyses. The BLS food composition table was used to ensure that all countries used the same food composition data (obtained via the same definitions and analytical methods), though it might also represent a limitation as not all country-specific recipes and foods could be found in this German food composition table. However, recipes were generated to calculate nutrient intakes for those particular recipes. The sex- and age-specific median-based cut-offs used are sample specific and, therefore, limits their comparability with other studies. Also, adolescents included in the present study cannot be considered a true representative sample due to differences in age, weight and height compared with the original HELENA sample.
The present study has several strengths. Blood samples were collected following a standardised methodology and transported to a centralised laboratory in order to ensure the viability and stability of the samples( Reference Gonzalez-Gross, Breidenassel and Gomez-Martinez 22 ). Fieldworkers were trained and a manual of operation was developed to guarantee good clinical practice( Reference Gonzalez-Gross, Breidenassel and Gomez-Martinez 22 ). The present results were adjusted for multiple testing by the Bonferroni method, which is considered as a very conservative method.
In conclusion, the present study showed the associations between energy-adjusted macronutrient intake and serum lipid profile in adolescents. Fat intake was related to a better serum lipid profile while carbohydrate intake was observed to be associated with an adverse lipid profile. It is noteworthy that these associations differed according to body fat status and were consistent across the obesity definitions used. These findings emphasise the importance of considering body fat status when developing strategies to prevent the risk of CVD among adolescents since serum lipids and obesity are the major markers of CVD risk. More research is needed, preferably with a longitudinal design, to confirm these findings.
Acknowledgements
The authors gratefully acknowledge all participating children and adolescents, and their parents and teachers for their collaboration. The authors thank Petra Pickert and Anke Carstensen for their contribution to the laboratory work.
The authors take the sole responsibility for the content of this article. The content of this article reflects the views of the authors only, and the European Community is not liable for any use that may be made of the information contained therein. The HELENA study was financially supported by the European Community Sixth RTD Framework Programme (contract FOOD-CT-2005-007034). The European Community had no role in the design or analysis of the study or in the writing of this article. The present study was also supported by the Aragón's Regional Government (SBS, grant no. B079/08), by a grant from the Spanish Ministry of Science and Innovation (grants no. JCI-2010-07055 and AP2008-03806) and by the European Regional Development Fund (MICINN-FEDER).
The authors’ contributions were as follows: L. A. M., F. G., A. K., Y. M., D. M., K. W. and M. S. participated in the conception and design of the study; S. B.-S., I. H., M. C.-G., G. P., C. B., R. R., K. V., M. S., L. L. and S. G.-M. conducted the research and collected the data; S. B.-S. wrote the manuscript, analysed the data and generated the figures; T. M., I. H., I. L. and L. A. M. participated in the interpretation of the data; S. B.-S., T. M., I. H., I. L., M. C.-G., G. P., C. B., D. M., R. R., K. W., F. G., A. K., Y. M., K. V., M. S., L. L., S. G.-M. and L. A. M. critically reviewed the manuscript. All authors read and approved the final manuscript.
There are no conflicts of interest.
APPENDIX: HELENA Study Group
Co-ordinator: L. A. Moreno
Core Group members: L. A. Moreno, F. Gottrand, S. De Henauw, M. González-Gross, C. Gilbert
Steering Committee: A. Kafatos (President), L. A. Moreno, C. Libersa, S. De Henauw, J. Sánchez, F. Gottrand, M. Kersting, M. Sjöstrom, D. Molnár, M. González-Gross, J. Dallongeville, C. Gilbert, G. Hall, L. Maes, L. Scalfi
Project Manager: P. Meléndez
Universidad de Zaragoza (Spain): L. A. Moreno, J. Fleta, J. A. Casajús, G. Rodríguez, C. Tomás, M. I. Mesana, G. Vicente-Rodríguez, A. Villarroya, C. M. Gil, I. Ara, J. Revenga, C. Lachen, J. Fernández-Alvira, G. Bueno, A. Lázaro, O. Bueno, J. F. León, J. Mª. Garagorri, M. Bueno, J. P. Rey López, I. Iglesia, P. Velasco, S. Bel-Serrat, L. Gracia-Marco, T. Mouratidou, D. Jiménez-Pavón
Consejo Superior de Investigaciones Científicas (Spain): A. Marcos, J. Wärnberg, E. Nova, S. Gómez-Martínez, E. Ligia Díaz, J. Romeo, A. Veses, M. Angeles Puertollano, B. Zapatera, T. Pozo
Université de Lille 2 (France): L. Beghin, C. Libersa, F. Gottrand, C. Iliescu, J. Von Berlepsch
Research Institute of Child Nutrition Dortmund, Rheinische Friedrich-Wilhelms-Universität Bonn (Germany): M. Kersting, W. Sichert-Hellert, E. Koeppen
Pécsi Tudományegyetem (University of Pécs) (Hungary): D. Molnar, E. Erhardt, K. Csernus, K. Török, S. Bokor, Mrs Angster, E. Nagy, O. Kovács, J. Répasi
University of Crete School of Medicine (Greece): A. Kafatos, C. Codrington, M. Plada, A. Papadaki, K. Sarri, A. Viskadourou, C. Hatzis, M. Kiriakakis, G. Tsibinos, C. Vardavas, M. Sbokos, E. Protoyeraki, M. Fasoulaki
Institut für Ernährungs- und Lebensmittelwissenschaften – Ernährungphysiologie. Rheinische Friedrich Wilhelms Universität (Germany): P. Stehle, K. Pietrzik, M. González-Gross, C. Breidenassel, A. Spinneker, J. Al-Tahan, M. Segoviano, A. Berchtold, C. Bierschbach, E. Blatzheim, A. Schuch, P. Pickert
University of Granada (Spain): M. J. Castillo, A. Gutiérrez, F. B. Ortega, J. R. Ruiz, E. G. Artero, V. España, D. Jiménez-Pavón, P. Chillón, C. Sánchez-Muñoz, M. Cuenca
Istituto Nazionale di Ricerca per gli Alimenti e la Nutrizione (Italy): D. Arcella, E. Azzini, E. Barrison, N. Bevilacqua, P. Buonocore, G. Catasta, L. Censi, D. Ciarapica, P. D'Acapito, M. Ferrari, M. Galfo, C. Le Donne, C. Leclercq, G. Maiani, B. Mauro, L. Mistura, A. Pasquali, R. Piccinelli, A. Polito, R. Spada, S. Sette, M. Zaccaria
University of Napoli “Federico II” Dept of Food Science (Italy): L. Scalfi, P. Vitaglione, C. Montagnese
Ghent University (Belgium): I. De Bourdeaudhuij, S. De Henauw, T. De Vriendt, L. Maes, C. Matthys, C. Vereecken, M. de Maeyer, C. Ottevaere, I. Huybrechts
Medical University of Vienna (Austria): K. Widhalm, K. Phillipp, S. Dietrich, B. Kubelka, M. Boriss-Riedl
Harokopio University (Greece): Y. Manios, E. Grammatikaki, Z. Bouloubasi, T. Louisa Cook, S. Eleutheriou, O. Consta, G. Moschonis, I. Katsaroli, G. Kraniou, S. Papoutsou, D. Keke, I. Petraki, E. Bellou, S. Tanagra, K. Kallianoti, D. Argyropoulou, K. Kondaki, S. Tsikrika, C. Karaiskos
Institut Pasteur de Lille (France): J. Dallongeville, A. Meirhaeghe
Karolinska Institutet (Sweden): M. Sjöström, J. Ruiz, F. B. Ortega, M. Hagströmer, L. Hallström, E. Patterson, L. Kwak, J. Wärnberg, N. Rizzo, A. Hurtig Wennlöf
Asociación de Investigación de la Industria Agroalimentaria (Spain): J. Sánchez-Molero, E. Picó, M. Navarro, B. Viadel, J. E. Carreres, G. Merino, R. Sanjuán, M. Lorente, M. J. Sánchez, S. Castelló
Campden BRI (United Kingdom): C. Gilbert, S. Thomas, E. Allchurch, P. Burguess
SIK – Institutet foer Livsmedel och Bioteknik (Sweden): G. Hall, A. Astrom, A. Sverkén, A. Broberg
Meurice Recherche & Development asbl (Belgium): A. Masson, C. Lehoux, P. Brabant, P. Pate, L. Fontaine
Campden and Chorleywood Food Development Institute (Hungary): A. Sebok, T. Kuti, A. Hegyi
Productos Aditivos SA (Spain): C. Maldonado, A. Llorente
Cárnicas Serrano SL (Spain): E. García
Cederroth International AB (Sweden): H. von Fircks, M. Lilja Hallberg, M. Messerer
Lantmännen Food R&D (Sweden): M. Larsson, H. Fredriksson, V. Adamsson, I. Börjesson
European Food Information Council (Belgium): L. Fernández, L. Smillie, J. Wills
Universidad Politécnica de Madrid (Spain): M. González-Gross, J. Valtueña, U. Albers, R. Pedrero, A. Meléndez, P. J. Benito, D. Cañada, D. Jiménez-Pavón, A. Urzanqui, J. C. Ortiz, F. Fuentes, J. J. Gómez Lorente, R. Mardía Torres, P. Navarro
