The prevalence of obesity has increased all over the world(Reference Bray, Paeratakul and Popkin1). Obesity may lead to morbidity such as hypertension(Reference MacCahon, Cutler, Brittain and Higgins2) and type 2 diabetes(Reference Colditz, Willet, Stampfer, Manson, Hennekens, Arky and Speizer3), and premature mortality(Reference Pedone, Urbinati, Pallotti and Pinelle4, Reference Lew and Garfinkel5). Some authors have stated that dietary fat can contribute to obesity via passive over-consumption, because this macronutrient is less satiating than either carbohydrates or proteins(Reference Poppitt and Prentice6) and is the most energy-dense macronutrient(Reference Poppitt and Prentice6, Reference Swinburn, Caterson, Seidell and James7). Recently, attention has been drawn to the type of fatty acids in the diet because of their differential metabolism, which is explained mostly by their chain length, saturation degree and stereoisomeric configuration(Reference Jones and Schoeller8–Reference Dulloo, Fathi, Mensi and Girardier11). Some investigators have proposed that dietary fat composition, independently of the amount of fat intake, can affect the development of obesity(Reference Jones and Schoeller8). Authors have suggested that short- and medium-chain fatty acids have a higher oxidation rate(Reference Jones and Schoeller8–Reference Dulloo, Fathi, Mensi and Girardier11) and may prevent obesity(Reference Dulloo, Fathi, Mensi and Girardier11–Reference Geliebter, Torbay, Bracco, Hashim and Van Itallie13). Others have reported that dietary MUFA(Reference Jones, Pencharz and Clandinin14, Reference Kien, Bunn and Ugrasbul15), particularly oleic acid such as found in olive oil(Reference Schröder16), and PUFA(Reference Leyton, Drury and Crawford17–Reference Flachs, Horakova, Brauner, Rossmeisl, Pecina and Franssen-van19), especially those found in fish oil, may promote weight loss. Although outcomes are not always consistent(Reference Awad, Bernardis and Fink20–Reference Kafatos, Diacaton, Voukiklaris, Nikolalakis, Vlachonilkolis, Kounali, Mamalakis and Donatas22), these results are reported mostly in animal studies and in a few clinical trials. Few studies in epidemiology concerning this issue are currently available and the results are conflicting(Reference Kafatos, Diacaton, Voukiklaris, Nikolalakis, Vlachonilkolis, Kounali, Mamalakis and Donatas22–Reference Bes-Rastrollo, Sanchez-Villegas, de la Fuente, de Irala, Martinez and Martinez-Gonzalez29). To examine the general trend in the world on the relationship between type of fat available for human consumption and obesity, an ecological study was conducted. The purpose of the study was to test for the existence of an association between obesity prevalence and types of fat available in 168 countries.
Methods
An ecological study of 168 countries was conducted. Data on the prevalence of obesity (BMI ≥ 30 kg/m2) among women aged 15 years and over were obtained from the WHO Global InfoBase(30). Food balance sheets (FBS) for the years 1998 to 2002 were obtained from the FAOSTAT database(31). Five-year averages for energy, total fat, MUFA, PUFA, SFA and ‘other fat’ per capita were calculated. The category ‘other fat’ is a category in the FBS regrouping all oils that have not been listed separately as other items. Their fatty acids content cannot therefore be estimated. We used the US Department of Agriculture Nutrient Data Laboratory database(32) and the Canadian Nutrient File(33) to derive fatty acids from the types of fat available for human consumption in each country. One hundred and sixty-eight countries were selected according to the availability of FBS in the FAOSTAT database and BMI percentage in the WHO website. According to FAOSTAT, the FBS presents a comprehensive picture of a country’s food supply pattern during a specific period. For each food item the FBS shows what is potentially available for human consumption, referring to the sources of supply and utilisation. Furthermore, the FAOSTAT database also gives the per capita supply of each food item available for human consumption, obtained by dividing the respective quantity by the population actually consuming it. These ‘per capita’ figures refer to one-year availability of food supply. All data are presented as means with standard deviation. Spearman correlations between obesity prevalence and potential predictors (energy, total fat, PUFA, MUFA, SFA and ‘other fat’ in g/capita per d) were performed. To elucidate the relationship between the type of fat available in these countries and obesity prevalence we conducted multiple linear regression models. Statistical significance was accepted at the 5 % level. All analyses were performed using the Statistical Analysis Systems statistical software package version 8 (SAS Institute, Cary, NC, USA).
Results
The characteristics of the various countries are presented in the Appendix. The prevalence of obesity ranged from 0 % in Ethiopia to 49·2 % in Kuwait. There was a wide variation in total fat consumption, from 10·5 g/capita in Burundi to 159·1 g/capita in Belgium. Means and standard deviations for the variables studied, together with Spearman correlation coefficients for the association of dietary variables with the prevalence of obesity, are presented in Table 1. Significant positive correlations were observed between obesity prevalence and energy (0·48), total fat (0·51), MUFA (0·41), PUFA (0·43), SFA (0·45), and ‘other fat’ (0·41). Furthermore, the types of fat were correlated positively with each other, with energy and total fat, and all results were statistically significant. The contribution of each fat group is also presented in Table 1 as a percentage of total energy intake, since recommendations are often reported in such terms. Similarly to the absolute contributions, the percentage contribution of each type of fat also increased in countries with higher obesity prevalence, but the correlations, although all still significant, were weakened slightly.
Table 1 Mean and standard deviation of variables studied in 168 countries and Spearman correlations (ρ) between obesity prevalence and energy, total fat, SFA, MUFA, PUFA and ‘other fat’
*Correlations were considered significant at P < 0·05.
We conducted multiple linear regression analyses to separate the relationships of each type of fat with obesity prevalence controlling for per capita energy intake. Note that the sum of all four types of fat (SFA, MUFA, PUFA and ‘other’) equals the total fat per capita and therefore this last variable was not included in the model. As expected, SFA (β = 0·38, P < 0·0001), PUFA (β = 0·68, P < 0·0001) and ‘other fat’ (β = 0·44, P = 0·02) were significantly positively associated with obesity. However, we found a significant negative association (β = −0·68, P < 0·0001) between MUFA availability and the prevalence of obesity (Table 2).
Table 2 Results of multiple linear regression analyses of dietary variables v. obesity prevalence (percentage of women in the population with BMI ≥ 30 kg/m2) as dependent variable in 168 countries
R 2 = 0·32.
*Correlations were considered significant at P < 0·05.
Discussion
The main result of the present paper is that, in spite of the significant positive association between obesity prevalence and total fat availability, MUFA availability is significantly negatively associated with the prevalence of obesity. It suggests that populations with lower obesity prevalence seem to consume greater amounts of MUFA, but such association cannot be taken as causal with our ecological study design. Nevertheless, this finding supports results from a few epidemiological studies reporting that the Mediterranean diet seems to be beneficial to weight loss(Reference Schröder16, Reference Shubair, McColl and Hanning34, Reference Panagiotakos, Chrysohoou, Pitsavos and Stefanadis35). In these studies, the authors specifically considered the consumption of olive oil, and not all types of MUFA in the diet. In contrast, other studies have reported that olive oil or the Mediterranean diet may promote weight gain(Reference Lovejoy, Smith, Champagne, Most, Lefevre, DeLany, Denkins, Rood, Veldhuis and Bray21, 24, Reference Gonzalez, Pera and Quiros25). Yet other investigators have not shown any relationship between a high consumption of MUFA and the prevalence of obesity(Reference Bes-Rastrollo, Sanchez-Villegas, de la Fuente, de Irala, Martinez and Martinez-Gonzalez29, Reference Trichopoulou, Gnardellis, Benetou, Lagiou, Bamia and Trichopoulos36, Reference Trichopoulou, Naska, Orfanos and Trichopoulos37). Some clinical trials(Reference Kien, Bunn and Ugrasbul15, Reference Soares, Cummings, Mamo, Kenrick and Piers38–Reference Ros41) but not all(Reference Lovejoy, Smith, Champagne, Most, Lefevre, DeLany, Denkins, Rood, Veldhuis and Bray21) have demonstrated that MUFA have a higher oxidation rate than SFA. In fact, the mechanism underlying this negative relationship, according to these studies, is that MUFA intake increases diet thermogenesis, which in turn stimulates the sympathetic nervous system(Reference Jones, Pencharz and Clandinin39), and abdominally obese subjects may be more responsive to stimulation of the sympathetic nervous system because they have an increased density and sensitivity of β-adrenoreceptors(Reference Bouchard, Despres and Mauriege42). Similarly, some studies(Reference Guzman, Lo Verme, Fu, Oveisi, Blazquez and Piomelli43) in mice demonstrated that MUFA consumption might have an anti-obesity action. These authors reported that MUFA intake may stimulate fat utilisation through activation of the nuclear receptor, PPAR-α. Others(Reference Tsunoda, Ikemoto, Takahashi, Maruyama, Watanabe, Goto and Ezaki44) have demonstrated that rats with a high MUFA intake may gain weight.
Our multivariate model also suggests that, in countries with higher prevalence of obesity, dietary MUFA tend to give place to some SFA and more so to PUFA consumption. In fact, it has been reported that a high PUFA intake may promote weight gain(Reference Gonzalez, Pera and Quiros25, Reference Brunner, Wunsuch and Marmot27). When comparing eighty-eight children from Crete and Cyprus, two Mediterranean islands, regarding the association of adipose tissue arachidonic acid content with BMI and overweight status, Savva et al.(Reference Savva, Chadjigeorgiou, Hatzis, Kyriakakis, Tsimbinos, Tornaritis and Kafatos45) found higher mean levels of arachidonic acid, dihomo-γ-linolenic acid and DHA in overweight and obese children. A positive association between adipose tissue arachidonic acid and BMI was noted. On the other hand, Ailhaud et al.(Reference Ailhaud, Massiera, Weill, Legrand, Alessandri and Guesnet46) reported that the inclusion of α-linolenic acid coming from PUFA in an isoenergetic diet rich in linoleic acid prevents increase of fat mass in pups. The authors highlighted that these data were consistent with their previous in vitro results comparing the adipogenic effect of n-6 PUFA and n-3 PUFA. Concerning SFA consumption and weight change, Doucet et al.(Reference Doucet, Alméras, White, Despres, Bouchard and Tremblay23) and Gonzalez et al.(Reference Gonzalez, Pera and Quiros25) reported a higher consumption of SFA in obese populations. Furthermore, some clinical trials(Reference Kien, Bunn and Ugrasbul15, Reference Piers, Walker, Stoney, Soares and O’Dea40) have demonstrated a higher oxidation rate in subjects who were consuming MUFA than in a group with SFA intake, for an isoenergetic diet. Kien et al.(Reference Kien, Bunn and Ugrasbul15) suggested that a high SFA intake (palmitic acid) may increase the obesity rate. Sanders(Reference Sanders47) demonstrated that populations with higher MUFA consumption tend to have lower intake of SFA, but we did not find such an association at the ecological level.
The present study has some positive points. First, the data on obesity prevalence were derived for all countries from the same recent WHO data set(30). FBS were also derived from one online database, FAOSTAT. These FBS represent the pattern of a country’s food supply during one year. Moreover, according to FAOSTAT, the quantity of foodstuff produced in a country added to the total quantity imported and adjusted to any change in stocks during a period of time gives the availability of supply during that period. These tables provide a useful reference for fat consumption for all countries(31).
For the statistical analysis we carried out multiple linear regression analyses to adjust for energy and estimate the respective contribution of each group of fats. This model explains 32 % of the variance found in the prevalence of obesity.
However, there are some limitations. Obviously, we cannot assume a negative cause-and-effect relationship between MUFA intake and obesity prevalence because the potential bias of ecological fallacy is always possible. This relationship may be totally or partially confounded by other unmeasured variables such as physical activity, geographical situation, consumption of dietary fibre, and fruit and vegetable intake. We are conscious of the fact that the FBS gives the food supply availability for the entire population in a country but obesity percentages taken into account in the present paper only include women aged 15 years and over. Consequently this relationship might be different for men, but the prevalence of obesity among men and women in a country is probably highly correlated. Another potential limitation is utilisation of the FBS, which is an estimation of the food supply available for human consumption in a given country, and that the validity of national reports may vary from country to country. The potential consequences of these variations in our analysis cannot be estimated. Also, the ‘other fat’ category that we had to use must have added imprecision to our estimates. An associated bias is nevertheless unlikely since its absolute contribution is small and represents probably a variety of fats. Finally, the availability for human consumption of more specific types of fatty acids and the n-6:n-3 ratio could not be taken into consideration for statistical analysis in the present paper, because of the imprecision and missing values of some particular items in the FBS.
This is the first ecological study to consider the type of fat and the prevalence of obesity in a large data set of 168 countries, since data on obesity from the WHO became available only recently. Our analysis suggests that additional studies on the potential role of MUFA in obesity are needed. Future use of online data sets is also encouraged.
Acknowledgements
The authors declare that the present paper has not been considered for publication elsewhere. Furthermore, there are no conflicts of interest or sources of funding for writing and publishing the paper. There are no acknowledgements to make for the article.
Authors’ contributions: N.M., who is a PhD student, wrote the article, O.R. (director) and V.G. (co-director) corrected and reviewed the article.
Appendix – Characteristics of 168 countries
Total fat, MUFA, PUFA, SFA and other fat represent all the quantities of fat available for human consumption per capita.
*Obesity prevalence is the percentage of women with BMI ≥ 30 kg/m2 in the population of each country.
