Genes encoding factors regulating food and energy intake

It is generally accepted that hypothalamic and brain stem centres are involved in the regulation of food intake and energy balance but only in the last decade has information on the relevant regulatory factors and their genes become available⁽Reference Marti, De Miguel, Jebb, Lafontan, Laville, Palou, Remesar, Trayhurn and Martínez⁵⁾. Insulin was regarded as the only candidate for the key role in body-weight regulation until the discovery of leptin by Friedman and colleagues⁽Reference Zhang, Proenca, Maffei, Barone, Leopold and Friedman⁶⁾, one of the most exciting findings of the last decade⁽Reference Marti, Novo, Martínez-Anso, Zaratiegui, Aguado and Martínez^7, Reference Marti, Berraondo and Martínez⁸⁾. This cytokine-like peptide, which is mainly expressed by adipocytes, is now believed to be a key regulator of fat metabolism and energy intake together with other adipokines⁽Reference Fischer-Posovszky, Wabitsch and Hochberg⁹⁾.

Certain areas of the hypothalamus are rich in specific receptors binding regulatory peptides and triggering central regulatory mechanisms. Factors acting at the central nervous system level include neuropeptide Y, corticotrophin-releasing hormone, pro-opiomelanocortin, α-melanocyte-stimulating hormone, agouti-related protein, melanin-concentrating hormone and cocaine- and amphetamine-regulated transcript⁽Reference Kamiji and Inui¹⁰⁾. Interactions between these molecules involving complex neuronal mechanisms eventually influence behaviour and provide important links with neuroendocrine regulation of other vital functions of the organism⁽Reference Walley, Blakemore and Froguel¹¹⁾.

Evidence is accumulating that most of the genes encoding central peptide factors as well as their receptors (leptin receptors, melanocortin receptors, neuropeptide Y receptors) are polymorphic⁽Reference Butler¹²⁾. Dominant inheritance of obesity conferred by missense, nonsense and frameshift mutations in the melanocortin 4 receptor (MC4R) gene has been extensively reported in many populations, including Spanish individuals⁽Reference Marti, Corbalán, Forga, Martínez, Hinney and Hebebrand^13, Reference Ochoa, Azcona and Biebermann¹⁴⁾. It has been estimated that 1–6% of extremely-obese individuals harbour functionally-relevant MC4R mutations⁽Reference Govaerts, Srinivasan, Shapiro, Zhang, Picard, Clement, Lubrano-Berthelier and Vaisse¹⁵⁾. More than seventy mutations of MC4R (fifty-seven non-synonymous, five nonsense and ten frameshift mutations) have been reported, many of them associated with dominant inheritance of obesity⁽Reference Ochoa, Azcona and Biebermann¹⁴⁾. Functional studies showed that many of the missense mutations also lead to a loss of function of MC4R⁽Reference Govaerts, Srinivasan, Shapiro, Zhang, Picard, Clement, Lubrano-Berthelier and Vaisse¹⁵⁾. Meanwhile, other mutations (i.e. Thr11Ser, Arg18Cys) and two polymorphisms (Val103Ile, Ile251Leu) do not modify the function of the MC4R in vitro ⁽Reference Ochoa, Azcona and Biebermann¹⁴⁾.

A number of peptides synthesised along the gastrointestinal tract also affect food intake. They include ghrelin (an orexigenic peptide mainly produced in the stomach), cholecystokinin (produced in the small intestine and acting as a short-term satiety signal) and peptide YY3–36 (produced in the colon and suppressing appetite for ⩽12 h)⁽Reference Coll, Farooqi and O'Rahilly¹⁶⁾. Exploration of these signalling pathways has started and it is becoming clear that polymorphism in relevant genes may have important functional consequences. For the ghrelin receptor gene two single-nucleotide polymorphisms have been reported: Ala204Glu and Phe279Leu, which selectively impair the constitutive activity of the receptor in human subjects leading to short stature and obesity that apparently develops during puberty⁽Reference Higgins, Gueorguiev and Korbonits¹⁷⁾.

Genes encoding factors implicated in energy expenditure

Adaptive thermogenesis in human subjects is closely related to the active mobilisation of lipids from fat tissues and is of particular interest in relation to obesity⁽Reference Dulloo, Seydoux and Jacquet¹⁸⁾. Central neural pathways responsible for food-intake and energy-expenditure regulation are closely interconnected. The peripheral transmission of central commands to the fat stores is mediated by the sympathetic nervous system. The β-adrenoceptor (ADRB) gene family members (ADRB2, ADRB3, ADRB1) are extensively-studied candidate genes in the obesity field because of their participation in energy-expenditure regulation⁽Reference Ochoa, Marti and Martinez¹⁹⁾.

The ADRB2 gene encodes a major lipolytic receptor protein in human fat cells. Two common polymorphisms of the ADRB2 gene, characterised by an amino acid replacement of arginine by glycine in codon 16 (Arg16Gly) and glutamine by glutamic acid in codon 27 (Gln27Glu) have been explored in several diseases, e.g. hypertension and obesity⁽Reference Macho-Azcárate, Calabuig, Marti and Martínez²⁰–Reference Macho-Azcárate, Marti, Calabuig and Martínez²²⁾. A relationship between the Arg16Gly polymorphism and an altered function of ADBR2 has been reported, leading to decreased agonist sensitivity⁽Reference Large, Hellström, Reynisdottir, Lönnqvist, Eriksson, Lannfelt and Arner²³⁾. Meanwhile, the Gln27Glu variant has also been found to be linked to obesity in some populations. In men the 27Glu allele has been associated with increased BMI and subcutaneous fat and with elevated leptin and TAG levels, while in women the 27Glu variant has been reported to be linked to increased BMI, body fat mass and waist:hip ratio⁽Reference Corbalán, Marti, Forga, Martínez-González and Martínez²⁴⁾.

The ADRB3 protein plays a role in adipocyte metabolism, mediating the rate of lipolysis in response to catecholamines, and ADRB3 agonists have potential anti-diabetes and anti-obesity properties⁽Reference Ochoa, Marti and Martinez¹⁹⁾. A common polymorphism in this gene, characterised by an amino acid replacement of tryptophan by arginine at position 64 (Trp64Arg), has been identified and may be linked to lower lipolytic activity and may account for lipid accumulation in the adipose tissue⁽Reference Fujisawa, Ikegami, Kawaguchi and Ogihara²⁵⁾. A number of reports have indicated a relationship between the Trp64Arg variant of ADRB3 and obesity-related phenotypes⁽Reference Fujisawa, Ikegami, Kawaguchi and Ogihara²⁵⁾. In relation to BMI, more than nine studies had shown an association between BMI and the Trp64Arg polymorphism in populations varying from 134 to 856 subjects⁽Reference Fujisawa, Ikegami, Kawaguchi and Ogihara²⁵⁾. In addition, two meta-analyses examining the effect of this mutation on BMI have been published for Caucasian populations⁽Reference Fujisawa, Ikegami, Kawaguchi and Ogihara^25, Reference Corbalán, Marti, Forga, Martínez-González and Martínez²⁶⁾. One of these meta-analyses included 2447 subjects and the summary weighted mean difference in BMI was 0·30 (95% CI 0·13, 0·47) kg/m², indicating that variant carriers exhibited higher BMI (on the average, 0·30 kg/m² higher) than normal homozygous subjects⁽Reference Fujisawa, Ikegami, Kawaguchi and Ogihara²⁵⁾. The second involved 7399 subjects but the results indicated no association⁽Reference Corbalán, Marti, Forga, Martínez-González and Martínez²⁶⁾. The Trp64Arg polymorphism has been associated with abdominal or visceral fat obesity in several populations such as Caucasians and Japanese subjects⁽Reference Martínez, Enriquez, Moreno-Aliaga and Marti²⁾. Similarly, several studies carried out among Mexican American, Japanese and Caucasians women have shown that carriers of the Arg allele have a higher BMI and lower reduction in visceral fat after weight loss⁽Reference Marti and Martínez³⁾. Interestingly, a gene–gene interaction between the Trp64Arg variant of the ADRB3 gene and the Pro12Ala variant of the PPARγ2 gene has been reported in Spanish and Mexican populations⁽Reference Ochoa, Marti, Azcona, Chueca, Oyarzábal, Pelach, Patiño, Moreno-Aliaga, Martínez-González and Martínez²⁷⁾.

Whereas ADRB participate in the regulation of adaptive thermogenesis as a component of sympathetic responses, uncoupling proteins (UCP) are involved in the modulation of heat-generating uncoupled respiration at the mitochondrial level⁽Reference Mozo, Emre, Bouillaud, Ricquier and Criscuolo²⁸⁾. They represent a family of carrier proteins localised in the inner layer of mitochondrial membranes. There are different members: UCP 1 is mostly expressed in brown adipose tissue and has a role in thermogenesis; UCP2 is ubiquitously present in all tissues; UCP3 is mainly expressed in skeletal muscle and brown adipose tissue⁽Reference Corbalán, Margareto, Martínez and Marti^29, Reference Marti, Larrarte, Novo, García, Warden and Martínez³⁰⁾. Their putative role as ‘uncoupling proteins’ has been intensively explored. Like UCP1, UCP2 mediates mitochondrial proton leak, releasing energy stores as heat and therefore affecting the efficiency of energy metabolism⁽Reference Larrarte, Novo, Marti and Martínez³¹⁾. The actual functions for UCP2 and UCP3 proteins are still under investigation. It has been proposed that UCP act as regulators of energy metabolism, being transmembrane fatty acid transporters in the mitochondria facilitating proton exchange⁽Reference Marti, Vaquerizo, Zulet, Moreno-Aliaga and Martínez^32, Reference Marti, Corbalán, Forga, Martínez-González and Martínez³³⁾.

Moreover, a number of human studies have indicated a relationship between UCP polymorphisms and exercise efficiency, resting energy expenditure, substrate oxidation, energy metabolism, BMI, obesity risk, type 2 diabetes risk, leptin, fat accumulation, body-weight changes, physical activity etc.⁽Reference Ochoa, Marti and Martinez¹⁹⁾. These observations have led to the consideration of UCP2 and UCP3 as candidate genes for obesity, given their function in the regulation of fuel metabolism⁽Reference Dalgaard and Pedersen³⁴⁾.

Several UCP2 gene variants have been described: a G/A mutation in the promoter region –866G/A, a valine for alanine substitution at amino acid 55 in exon 4 (Ala55Val) and a 45 bp insertion/deletion in the untranslated region of exon 8⁽Reference Marti, Corbalán, Forga, Martínez-González and Martínez³³⁾. The associations between these polymorphisms of UCP2 and various aspects of obesity have been intensively studied. From the literature, it seems that the G allele in the promoter region of UCP2 increases obesity risk, while affording relative protection for type 2 diabetes⁽Reference Marti, Corbalán, Forga, Martínez-González and Martínez³³⁾. On the other hand, the Ala55Val polymorphism has been shown to be associated with increased exercise efficiency⁽Reference Ochoa, Marti and Martinez¹⁹⁾. However, findings relating to the exon 8 insertion allele of the UCP2 gene have been inconsistent. While no association with obesity has been observed in a number of the studies conducted in several populations, significant associations have been found between the exon 8 insertion of the UCP2 gene and BMI, fat mass and the presence of obesity (ranging from P<0·01 to P<0·001)⁽Reference Ochoa, Marti and Martinez¹⁹⁾.

There are also several UCP3 gene variants. In linkage studies some of the variants have been shown to be associated with a higher obesity risk. Specifically, the –55C/T polymorphism in the promoter region of this gene has been shown to be associated with an elevated BMI, an increased level of adiposity or a greater waist:hip ratio⁽Reference Ochoa, Marti and Martinez¹⁹⁾. However, other authors have not found any relationship between this polymorphism and a higher risk of obesity or changes in metabolic rate⁽Reference Ochoa, Santos, Azcona, Moreno-Aliaga, Martínez-González, Martínez and Marti³⁵⁾. Some studies have even reported an inverse correlation with BMI and the presence of the –55C/T polymorphism⁽Reference Alonso, Martí, Corbalán, Martínez-González and Martínez³⁶⁾.

As the UCP2–UCP3 gene cluster extends within a small region of 40 kb of the genome, a haplotype study is a useful tool to study its association with obesity. Using a study sample comprising 193 obese children and adolescents (cases) and 170 controls aged 6–18 years it has been found that the individual polymorphisms are not associated with obesity, but the (–866G; rs659 366)–(Del; 45 bp)–(–55T; rs1 800 849) haplotype is significantly associated with obesity and its presence in the control group increases the insulin resistance risk by about nine times⁽Reference Ochoa, Santos, Azcona, Moreno-Aliaga, Martínez-González, Martínez and Marti⁴⁾.

Genes encoding factors implicated in adipogenesis

The last group of genes acting in connection with peripheral regulation of energy expenditure comprises the transcription factors leading to adipogenesis and adipocyte differentiation. The key factors are the PPARγ, particularly the adipose-specific isoform PPARγ2⁽Reference Margareto, Marti and Martínez^37, Reference López, Marti, Milagro, Zulet, Moreno-Aliaga and Martínez³⁸⁾. In a meta-analysis examining the Pro12Ala polymorphism in 19 136 subjects a positive association with BMI has been found⁽Reference Marti and Martínez³⁾. The frequency of the Ala allele, similar to other Caucasian populations, has been found to be higher in obese subjects (allelic frequency 0·13) than in controls (allelic frequency 0·08), suggesting that this polymorphism is associated with obesity⁽Reference Ochoa, Marti and Martinez^19, Reference Marti, Corbalán, Martínez-González, Forga and Martínez³⁹⁾. There is also information on the functional role of PPARγ gene variants: some chimeric proteins appear to have a reduced activity⁽Reference Stumvoll and Häring⁴⁰⁾.

Gene variants, carbohydrate consumption and obesity risk

Few attempts have been made to examine the influence of gene–diet interactions on obesity-related phenotypes. A case–control design with selected criteria for inclusion (BMI >30 kg/m² for cases and BMI <25 kg/m² for controls) has found a significant interaction between obesity risk and high levels of CHO consumption (>49% total energy) for carriers of the Glu allele of the ADRB2 gene⁽Reference Martínez, Corbalán, Sánchez Villegas, Forga, Marti and Martínez-González⁴⁴). Obesity incidence was not found to be directly affected by the polymorphism (OR 1·40, P=0·246), but using a multivariate logistic model after adjustment for confounding factors and/or effect modifiers a marginally-significant interaction between the single-nucleotide polymorphism and CHO intake was found among women. This finding suggests that a dietary intake higher than the median CHO consumption (>49% total energy) produces an increased obesity risk in those women carrying the Glu27 allele (OR 2·56, P=0·051) when adjusted for age and physical activity. A second model, using CHO:fat instead of the percentage energy derived from CHO, supports these results; a CHO intake:fat intake of >1·77 was found to be associated with a higher risk of obesity (OR 3·21, P<0·02) in women with the ADRB2 gene polymorphism. The natural logarithms of the OR of being obese for women with and without the polymorphism according to the intake of CHO were found to be marginally significant for women with the polymorphism (b 0·074, P=0·056), using the lower level of intake as the reference and adjusting for age and physical activity (Fig. 1). Among female subjects bearing the Glu27 polymorphism an association between high CHO intake and higher insulin levels (P<0·03) was found, which may contribute to the greater likelihood of an onset of obesity among these subjects. The fact that some studies have shown that a polymorphism on the ADRB2 gene adversely reduces fat oxidation⁽Reference Macho-Azcarate, Marti, Calabuig and Martinez⁴⁵⁾, which may consequently be associated with hyperlipidaemia, insulin resistance and hyperinsulinaemia, also helps to explain that carriers of the Glu27 allele may show an impaired response after high CHO intake, leading to obesity.

Fig. 1. Obesity risk linked to the single-nucleotide polymorphism Gln27Glu of the β-adrenoceptor 2 gene depends on carbohydrate (CHO) consumption. Natural logarithms of the OR (LnOR) of being obese for women with (□) and without (▲) the polymorphism (logistic regression model) according to the intake of CHO (% energy) and adjusting for age and physical activity during leisure time (metabolic equivalent-hours/week). For interaction, P=0·056. (From Martínez et al.⁽Reference Martínez, Corbalán, Sánchez Villegas, Forga, Marti and Martínez-González⁴⁴⁾.)

Interestingly, a series of research papers has been devoted to the interplay between the polymorphism Pro12Ala of the PPARγ2 gene and dietary patterns in relation to obesity phenotypes⁽Reference Marti and Martínez³⁾. In vivo ligands for PPARγ2 are thought to include a variety of fatty acids according to their chain length or extent of saturation. An inverse interaction between dietary polyunsaturated fat:saturated fat and BMI among 12Ala carriers was found, the mean BMI being greater in Ala carriers than in Pro/Pro homozygotes when polyunsaturated fat:saturated fat is low. In a population of women of larger body size the intake of saturated fat was found to be directly associated with increased BMI in carriers and non-carriers of the 12Ala variant, whereas the intake of monounsaturated fat was shown to be inversely associated with BMI only in 12Ala carriers of the PPARγ2 polymorphism⁽Reference Memisoglu, Hu, Hankinson, Manson, De Vivo, Willett and Hunter⁴⁶⁾.

A case–control study has reported a higher risk of obesity for carriers of the 12Ala variant with increasing intake of arachidonic acid⁽Reference Nieters, Becker and Linseisen⁴⁷⁾, whereas the author's group has found an increased obesity risk for carriers of the 12Ala allele when consuming >49% total energy from CHO (Table 2)⁽Reference Marti, Corbalán, Martínez-González, Forga and Martínez³⁹⁾.

Table 2. Obesity risk linked to the Pro12Ala polymorphism of the PPARγ2 gene depends on carbohydrate (CHO) consumption (from Marti et al.⁽Reference Marti, Corbalán, Martínez-González, Forga and Martínez³⁹⁾)

Gene variants, physical activity levels and obesity risk

To estimate physical activity levels face-to-face interviews were conducted with volunteers and validated questionnaires relating to their participation (number of hours per week) in different sports, exercises or physical activities (using the compendium of physical activities⁽Reference Ainsworth, Haskell, Leon, Jacobs, Montoye, Sallis and Paffenbarger⁴⁸⁾) were completed. The amount and intensity of leisure time were quantified by assigning metabolic equivalents (energy expended during each specific activity:RMR) to each activity⁽Reference Ochoa, Moreno-Aliaga, Martinez-Gonzalez and Martinez^42, Reference Marti, Corbalán, Martínez-González and Martínez⁴⁹⁾.

In a case–control study that assessed the role of the Trp64Arg mutation of the ADRB3 gene on the risk of developing obesity it was found that the effect of the ADRB3 mutation on obesity risk changes depending on the recreational physical activity levels⁽Reference Marti, Corbalán, Martínez-González and Martínez⁴⁹⁾. The metabolic equivalent-hours/week:time spent sitting down during leisure time (M/S) was used to estimate recreational energy expenditure. An effect modification across recreational physical activity strata was apparent, and a univariate OR of 2·98 (95% CI 1·00, 8·56) was found within the sedentary group (low physical activity; M/S<0·5) whereas no increment in risk was apparent for active subjects (high physical activity; M/S>0·5). When the association between the risk of obesity and the Trp64Arg mutation was adjusted for gender and age using multivariate logistic regression models and introducing a product-term (age×ADRB3 mutation), a borderline significant interaction (P=0·06) between recreational activity and the ADRB3 mutation was demonstrated (Table 3). Thus, carriers of the Trp64Arg mutation with low levels of physical activity have the highest risk of obesity after adjustment for age and gender.

Table 3. Obesity risk linked to the single-nucleotide polymorphism Trp64Arg of the β-adrenoceptor 3 (ADRB3) gene depends on physical activity (PA) levels (from Marti et al.⁽Reference Marti, Corbalán, Martínez-González and Martínez⁴⁹⁾)

For interaction recreational activity×ADRB3 mutation, P=0·06.

^M/S, metabolic equivalent-hours per week: time spent sitting down during leisure time.

*

The association between the 27Glu polymorphism and obesity risk has been estimated using multivariate logistic regression⁽Reference Corbalán, Marti, Forga, Martínez-González and Martínez⁵⁰⁾. An effect modification (interaction) on the risk of obesity linked to the 27Glu polymorphism by the level of recreational energy expenditure (M/S) was derived after adjustment for age. A significant interaction (product term M/S×27Glu allele; P=0·005) between recreational energy expenditure and the 27Glu allele was demonstrated (Fig. 2). The mean BMI of the two groups was compared on the basis of energy expenditure using the 75th percentile of energy expenditure during leisure time as a cut-off (M/S 0·9). Interestingly, a significant interaction was found in the linear model between the Glu27 allele and the M/S (P=0·003). Women who were more active in their leisure time (M/S>0·9) and were carriers of the 27Glu allele had a higher BMI compared with non-carriers. The data demonstrate that obese women who are carriers of the 27Glu allele do not benefit equally from physical activity compared with non-carriers. Such individuals may be more resistant to losing weight when they participate in higher physical activity levels.

Fig. 2. Obesity risk linked to the single-nucleotide polymorphism Gln27Glu of the β-adrenoceptor 2 gene depends on physical activity levels. (A) The change in the magnitude of the association between the Gln27 allele (•, Gln27; ■, Glu27) and the obesity risk is dependent on the exposure to physical activity (metabolic equivalent-hours per week:time spent sitting down during leisure time; M/S). (B) Average BMI for subjects with (///) and without (□) the Glu27 polymorphism. Values are means with their standard errors represented by vertical bars. The table shows coefficients obtained with the multivariate logistic regression model using obesity (BMI>30 kg/m²) as the outcome and represent the independent effects for recreational energy expenditure (M/S), age and the Glu27 polymorphism and a product term assessing the effect modification of the polymorphism by M/S. (From Corbalán et al.⁽Reference Corbalán, Marti, Forga, Martínez-González and Martínez⁵⁰⁾.)

Physical activity in children and adolescents seems to be declining, while their time spent in sedentary activities such as television (TV) watching is increasing. Time spent watching TV replaces more vigorous activities and increases the likelihood of children adopting unhealthy food habits, thus it is a risk factor for obesity. This change in activity may explain why TV viewing is a significant predictor of BMI and overweight in childhood and has been validated as a good index of sedentary behaviour.

A matched case–control study was conducted to assess the interaction between the Gln27Glu polymorphism of the ADRB2 and TV viewing (as a proxy for sedentary lifestyles) in relation to obesity in a group of Spanish children and adolescents⁽Reference Ochoa, Moreno-Aliaga, Martínez-González, Martínez and Marti⁵¹⁾. The study population comprised 330 Spanish children and adolescents. Cases (n 165) were subjects aged 5–18 years with a BMI of >97th percentile of the Spanish BMI reference data for age and gender⁽Reference Sobradillo, Aguirre and Aresti⁵²⁾. Using logistic regression analysis it was found that the Glu27Glu genotype of the ADRB2 gene was not associated with obesity in boys in the fully-adjusted model. However, female subjects carriers of the 27Glu allele (twenty-five homozygous Glu27Glu and ninety-three heterozygous: Glu27Gln) were found to have a significantly higher risk of obesity (OR 1·95 (95% CI 1·02, 3·70)) in the completely-adjusted model. When the OR for obesity for girls were calculated for the presence of the 27Glu allele in homozygotes, the values were higher (OR 2·83 (95% CI 1·12, 7·19). In the fully-adjusted model when physical activity (metabolic equivalent-hours per week) and the usual time spent in watching TV were also taken into account and introduced in the model the adjusted OR for obesity linked to the genotype Glu27Glu in the female population rose to 4·84 (95% CI 1·37, 17·10).

An effect modification (interaction) on obesity risk linked to the Gln27Glu polymorphism of the ADRB2 gene by the number of hours spent watching TV (P=0·023) was observed after adjustment for physical activity (Fig. 3). In order to calculate different OR for sedentary and non-sedentary subjects the female population was subdivided using the median of the number of hours spent watching TV (12·5 h/week). In homozygous subjects for the wild-type genotype the risk of obesity was found to increase seven times for sedentary girls (i.e. they spent >12·5 h/week watching TV; OR 7·27 (95% CI 1·42, 37·13)) as compared with non-sedentary girls (<12·5 h/week watching TV). Surprisingly, among subjects who were 27Glu carriers even girls with a low level of TV watching (<12·5 h/week) were found to have a high obesity risk (OR 4·60 (95% CI 1·01, 20·02)), which was not very different from the OR values for sedentary girls (>12·5 h/week watching TV; OR 6·05 (95% CI 1·31, 27·71)). These results suggest that carriers of the 27Glu allele of the ADRB2 gene may not benefit from a reduction in sedentary behaviour as much as the subjects who do not carry the polymorphism.

Fig. 3. Obesity risk linked to the single-nucleotide polymorphism Gln27Glu of the β-adrenoceptor 2 gene depends on the time spent watching television (TV) in Spanish children and adolescents aged 5–18 years with a BMI of >97th percentile of the Spanish BMI reference data for age and gender⁽Reference Sobradillo, Aguirre and Aresti⁵²⁾. MET, energy expended during each specific activity: RMR; (▲), Carriers; (•), non-carriers. (From Ochoa et al.⁽Reference Ochoa, Moreno-Aliaga, Martínez-González, Martínez and Marti⁵¹⁾.)