Original Communication Obesity and body mass index in Spain: the ...

European Journal of Clinical Nutrition (2001) 55, 782–785 ß 2001 Nature Publishing Group All rights reserved 0954–3007/01 $15.00

www.nature.com/ejcn

Original Communication Obesity and body mass index in Spain: the ‘single population’ theory revisited JR Banegas1*, JL Gutie´rrez-Fisac1, F Rodrı´guez-Artalejo1, JJ Cruz1, P Guallar1 and R Herruzo1 1

Department of Preventive Medicine, School of Medicine. Universidad Autońoma de Madrid, Madrid, Spain

Objective: The ‘single population’ theory suggests that the distribution of body mass index (BMI) moves up and down as a whole. We test directly whether this theory is valid among the adult population within one country over time, by examining the association between median BMI and the prevalence of obesity in the adult population of Spain’s 17 regions between 1987 and 1993. Design and setting: Ecological study using data from two national health surveys. Subjects and interventions: Self-reported weight and height for persons aged 20 – 64 y were drawn from representative Spanish-population samples from two similar National Health Surveys carried out in 1987 (n ¼ 20 705) and 1993 (n ¼ 15 490). Main outcome measures: Correlation and linear regression between the difference in median BMI and the difference in the prevalence of obesity (BMI 30 kg=m2) between 1987 and 1993. Results: Between 1987 and 1993 the change in median BMI was in general strongly correlated with the change in the prevalence of obesity (r ¼ 0.85; P < 0.001). The correlation is greater in women, persons over 45 y, and those with lower educational levels. Between 1987 and 1993 each unit of increase in median BMI is associated with an absolute increase of 5.1% (95% CI 3.3 – 6.8%) in the prevalence of obesity. Conclusions: The single population hypothesis holds true among the adult population within a whole country over time. This could be useful for monitoring and understanding the prevalence of obesity within a population. Descriptors: obesity; overweight; body mass index; prevalence; trends; single-population theory European Journal of Clinical Nutrition (2001) 55, 782–785

Introduction Many developed countries have a high and increasing prevalence of obesity (Seidell & Flegal, 1997) that threatens to become the foremost cause of chronic disease in the world (Pi-Sunyer, 1993; Grundy, 1998). The term ‘epidemic’ of obesity implies that obesity is a characteristic of populations, not only of individuals (Flegal, 1999). The socalled ‘single population’ theory suggests that the distribution of body mass index (BMI) moves up and down as a whole, thus the average BMI and the prevalence of obesity are associated (Rose & Day, 1990; Rose, 1992).

*Correspondence: JR Banegas, Departamento de Medicina Preventiva, Facultad de Medicina, Universidad Autońoma de Madrid, Avda Arzobispo Morcillo s=n, 28029 Madrid, Spain. E-mail: [email protected] Guarantor: JR Banegas. Contributors: JLG, FR, JJC, PG and RH contributed to the study design, analysis of the data, and revision of the paper. Received 30 October 2000; revised 20 February 2001; accepted 6 March 2001

Some authors have investigated time trends in the BMI and in the prevalence of obesity in several countries (Kuskowska-Wolk & Bergstrom, 1993a,b; Kuczmarski et al, 1994; Hoffmeister et al, 1994; Galuska et al, 1996; Seidell & Flegal, 1997). However, the single population theory has mostly been directly tested among centres in different countries at a single point in time (Rose & Day, 1990; Rose, 1992), or studied in school boys and very young men over time (Sorensen & Price, 1990; Price et al, 1991; Thomsen et al, 1999). While Rose & Day found that the theory held true cross-sectionally across diverse societies in individuals 20 – 59 y (Rose & Day, 1990), some studies of Danish young men examined at around 19 y showed in some birth cohorts a secular increase in prevalence of obesity without similar trends in median BMI (Sorensen & Price, 1990). A similar picture emerges from studies in Danish school boys aged 7 – 13 y (Thomsen et al, 1999). However, later studies in Danish young men showed increases in the prevalence of obesity accompanied by a slightly increasing median BMI in more recent cohorts (Sorensen et al, 1997). Thus, in this study we test whether the single population theory is valid among the adult population within one

Obesity and BMI in Spain JR Banegas et al

783

country over time, by examining the statistical association between median BMI and the prevalence of obesity in the adult population of Spain’s 17 regions between 1987 and 1993. This study could be important for monitoring and understanding the obesity epidemic within a population. Methods Self-reported height and weight for persons aged 20 – 64 y were taken from two similar National Health Surveys carried out in 1987 (n ¼ 20 705) and 1993 (n ¼ 15 490), which were described elsewhere (Gutie´rrez-Fisac et al, 1996). Briefly, the samples were random and representative of the Spanish population. The response rates to the questions on height and weight were 80% in 1987 and 85% in 1993. We performed correlation and linear regression between the difference in median BMI and the difference in the prevalence of obesity in Spain’s 17 regions between 1987 and 1993. Obesity was defined as BMI 30 kg=m2 (Seidell & Flegal, 1997). The median rather than the mean was used because BMI distribution was skewed. The correlation and regression coefficients were weighted by the sum of the sample size in each region in 1987 and 1993. Age distributions did not significantly change in most of Spain’s regions between 1987 and 1993, so no age-adjustment was made.

to 1993, with increasing BMI, the entire curve shifted slightly to the right. In the year 1987, the Pearson’s correlation coefficient between median BMI and prevalence of obesity across the regions was 0.76 (P ¼ 0.0004), and in year 1993 it was 0.84 (P < 0.001). Between 1987 and 1993 the median BMI increased by 0.50 kg=m2 (25.26 kg=m2 in 1993 to 24.76 kg=m2 in 1987) in men and 0.31 kg=m2 (23.53 kg=m2 in 1993 to 23.22 kg=m2 in 1987) in women. This was accompanied by an absolute increase in the prevalence of obesity of 1.8% (8.5% in 1993 to 6.7% in 1987) in men and 1.9% (9.1% in 1993 to 7.2% in 1987) in women. Figure 2 displays this

Results Distribution curves are presented in Figure 1. There was no double hump in either curve or suggestion that the distribution curve was composed of two distinctly different curves representing both a normal and a diseased population. Distribution curves both in 1987 and 1993 are essentially normal, with slight skewing to the upper values. From 1987

Figure 1

Figure 2 Relation between the difference in median body mass index (BMI) and the difference in the prevalence of obesity in the 17 regions of Spain between 1987 and 1993.

Distribution of body mass index in Spain in years 1987 and 1993. s.d.: standard deviation. European Journal of Clinical Nutrition


784

information for the 17 regions studied. Regional variation in changes in median BMI and prevalence of obesity was slightly greater among women than among men (data not shown). Across the Spanish regions, the change in median BMI was in general strongly and positively associated with the change in the prevalence of obesity (Table 1). The correlation is greater in women, persons over 45 y and those with lower educational levels (Table 1). In general, the results did not change substantially when Spearman’s rather than Pearson’s coefficients (Table 1), or when the mean (without obese individuals contributing to the mean) rather than the median (r ¼ 0.66, P ¼ 0.004), were used. The correlation between changes in median BMI and prevalence of overweight (BMI ¼ 25 – 29.9 kg=m2) was also strong (r ¼ 0.81, P < 0.001). Finally, the correlation between changes in median BMI and very extreme obesity (BMI 35 kg=m2) was rather modest (the Pearson’s correlation coefficient was 0.36, P ¼ 0.15, and Spearman’s coefficient was 0.43, P ¼ 0.08). According to the regression results, each unit of increase in median BMI is associated with an absolute increase of 5.1% in the prevalence of obesity between 1987 and 1993 (Table 1). Discussion These results extend to an adult population within a whole country over time the evidence obtained at a single point in time in other studies (Rose & Day, 1990), showing that the single population theory holds true in Spain’s adult population between 1987 and 1993. The theory also held across the regions at each of the two surveys. The close link between median and prevalence suggests that the frequency of cases of obesity could be better monitored and understood in the context of the population’s BMI. Specifically, assessment of the median BMI may possibly be a feasible tool in monitoring the obesity epidemic in smaller population samples. Our results are also consistent with many studies that, though not testing directly the single population theory, have reported secular increase in the prevalence of obesity with increase in average BMI (Kuskowska-Wolk & Bergstrom, 1993a,b; Kuczmarski et al, 1994; Hoffmeister et al,

1994; Galuska et al, 1996). However, some studies have shown time increase in prevalence of obesity without similar trends in average BMI among some birth cohorts of young men and boys, which suggests changes in environmental influences operating at different ages and affecting different subsets of the population, either because of selective exposure or particular susceptibility (Sorensen & Price, 1990; Price et al, 1991; Thomsen et al, 1999). These findings have important implications, in that in young men prevalence of obesity could not be predicted or monitored from average BMI values. However, later studies in Danish young men showed increases in the prevalence of obesity and an increase also in the median BMI in more recent cohorts, which may call for alternative or supplementary hypotheses (Sorensen et al, 1997). In any case, these authors studied school boys aged 7 – 13 y and young men aged 18 – 20 y, whereas we studied adult men and women aged 20 – 64 y, and with a different methodology. On the other hand, BMI could be a less valid indicator of obesity in young populations due to the higher relative importance of lean mass, particularly in males. This may be one of the reasons for the lower correlation coefficient among changes in BMI median and prevalence of obesity we found in young men. For Spanish men aged 20 to 44-y old the correlation coefficient was 0.3 (P ¼ 0.2), while for women 20 to 44-y-old r ¼ 0.6 (P ¼ 0.01). Although obesity is based on BMI as well, the high cut-off to define obesity ( 30 kg=m2) assures a high probability of being obese. Thus, time changes in BMI median in young males may possibly reflect more changes in lean mass than in fat mass in this population. Another interesting finding is that the correlation between changes in median BMI and changes in the very extreme tail of the distributions (BMI 35 kg=m2) was rather modest. Perhaps the very obese people reflect to a greater extent the genetic influence on obesity, thus being less susceptible to increase with increases in BMI in the whole population. A variety of social factors acting on the whole population may underlie the association between median BMI and obesity (Grundy, 1998; James, 1995; Prentice, 1997; Weinsier et al, 1998). For example, the time spent in certain sedentary activities such as using a personal computer or watching television, the caloric contribution of fats, and eating in fast food restaurants increased substantially over

Table 1 Association between the difference in median BMI and the difference in the prevalence of obesity in the 17 regions of Spain between 1987 and 1993, by age, sex and educational level Age, sex and educational level Total < 45 y 45 y Men Women < 12 y of education 12 y of education a

Pearson’s correlation coefficient 0.85 0.34 0.82 0.67 0.82 0.81 0.67

(P < 0.001) (P ¼ 0.17) (P < 0.001) (P ¼ 0.003) (P < 0.001) (P < 0.001) (P ¼ 0.003)

Spearman’s correlation coefficient 0.84 0.78 0.78 0.74 0.83 0.78 0.62

(P < 0.001) (P < 0.001) (P < 0.001) (P ¼ 0.005) (P < 0.001) (P < 0.001) (P ¼ 0.008)

Absolute difference in prevalence of obesity (%) with 1 unit difference in median BMI between 1987 and 1993.

European Journal of Clinical Nutrition

Regression coefficient (95% CI)a 5.1 0.3 7.6 5.1 4.7 6.6 2.9

(3.3 – 6.8) ( 7 1.1 – 0.7) (4.7 – 10.4) (2.0 – 8.1) (2.9 – 6.5) (4.0 – 9.3) (1.1 – 4.6)


785

the last two decades in Spain (Instituto Nacional de Estadı´stica, 1994; Varela et al, 1995). In our study, the correlation between BMI and obesity was stronger in women, older persons and those with lower educational levels — the groups with the highest prevalences of both obesity and some related factors (Gutie´rrez-Fisac et al, 1996; Instituto Nacional de Estadı´stica, 1994; Varela et al, 1995). It may be that the groups in which the correlations are greater are those on which prevalences of obesity could be better predicted or monitored from the median BMI. These are also the groups watching most television in Spain, and the least well informed about a healthy diet and its relation to weight (Ministerio de Sanidad y Consumo, 1992; Regidor et al, 1995). Likewise, persons with a lower educational level have the highest calorie and fat consumption (Varela et al, 1995), and are the ones most influenced by advertising and fads, and many advertising activities target them specifically. These groups may require a more directed preventive strategy. In this study we used self-reported height and weight, which have relatively high validity (Nieto-Garcıá et al, 1990). Although BMI and obesity may have been underestimated, there is no evidence that this error would vary between the 1987 and 1993 surveys, since they used the same questionnaire on height and weight. Some conclusions can be drawn from the present study. In general the single-population hypothesis holds true in Spain’s adult population over time, which could be useful for monitoring the obesity epidemic. More specifically, the single population theory suggests that assessment of the average BMI in smaller population samples may possibly be a feasible tool in monitoring the obesity epidemic. However, some exceptions in younger males in our study, and findings in younger populations in other studies, call for further investigation. In addition, findings corroborating the theory create a research challenge in terms of understanding the determinants of the obesity epidemic: are there changes in determinants that affects the entire population that may be identified and modified? However, again exceptions or counter-examples among younger men call for further research in specific subgroups of the population.

References Flegal KM (1999): The obesity epidemic in children and adults: current evidence and research issues. Med. Sci. Sports Exerc. 31(Suppl 11), S509 – S514.

Galuska DA, Serdula M, Pamuk E, Siegel PZ & Byers T (1996): Trends in overweight among US adults from 1987 to 1993: a multistate telephone survey. Am. J. Public Health 86, 1729 – 1735. Grundy SM (1998): Multifactorial causation of obesity: implications for prevention. Am. J. Clin. Nutr. 67(Suppl), 563S – 572S. Gutie´rrez-Fisac JL, Regidor E & Rodrı´guez C (1996): Trends in obesity differences by educational level in Spain. J. Clin. Epidemiol. 49, 351 – 354. Hoffmeister H, Mensink GBM & Stolzenberg H (1994): National trends in risk factors for cardiovascular disease in Germany. Prev. Med. 23, 197 – 205. Instituto Nacional de Estadı´stica (1994): Panora´mica Social de Espanã. Madrid: INE. James WPT (1995): A public health approach to the problem of obesity. Int. J. Obes. Relat. Metab. Disord. 19(Suppl 3), S37 – S45. Kuczmarski RJ, Flegal KM, Campbell SM & Johnson CL (1994): Increasing prevalence of overweight among U.S. adults. The National Health and Nutrition Examination Surveys, 1960 to 1991. J.A.M.A. 272, 205 – 211. Kuskowska-Wolk A & Bergstrom R (1993a): Trends in body mass index and prevalence in Swedish men 1980 – 89. J. Epidemiol. Community Health 47, 103 – 108. Kuskowska-Wolk A & Bergstrom R (1993b). Trends in body mass index and prevalence in Swedish women 1980 – 89. J. Epidemiol. Community Health 47, 195 – 199. Ministerio de Sanidad y Consumo (1992): Estudio de los estilos de vida de la poblacioń adulta espanõla. Madrid: Ministerio de Sanidad. Nieto-Garcıá FJ, Bush TL & Kely PM (1990): Body mass definition of obesity, sensitivity and specificity using self-reported weight and height. Epidemiology 1, 146 – 152. Pi-Sunyer FX (1993): Medical aspects of obesity. Ann. Intern. Med. 119, 655 – 660. Prentice AM (1997): Obesity — the inevitable penalty of civilization? Br. Med. Bull. 53, 229 – 237. Price RA, Ness R & Sorensen TIA (1991): Changes in commingled body mass index distributions associated with secular trends in overweight among Danish young men. Am. J. Epidemiol. 133, 501 – 510. Regidor E, Gutie´rrez-Fisac JL, Rodrı´guez C, de Mateo S & Alonso I (1995): Las desigualdades sociales y la salud en Espanã. In SESPAS. La salud y el sistema sanitario en Espanã. Informe SESPAS 1995. Barcelona: SG Editores. Rose G (1992): The Strategy of Preventive Medicine. Oxford: Oxford University Press. Rose G & Day S (1990): The population mean predicts the number of deviant individuals. Br. Med. J. 301, 1031 – 1034. Seidell JC & Flegal KM (1997): Assessing obesity: classification and epidemiology. Br. Med. Bull. 53, 238 – 252. Sorensen TI & Price RA (1990): Secular trends in body mass index among Danish young men. Int. J. Obes. 14, 411 – 419. Sorensen HT, Sabroe S, Gillman M, Rothman KJ, Madsen KM, Fischer P & Sorensen TIA (1997): Continued increase in prevalence of obesity in Danish young men. Int. J. Obes. Relat. Metab. Disord. 21, 712 – 714. Thomsen BL, Ekstrom CT & Sorensen TIA (1999): Development of the obesity epidemic in Denmark: cohort, time and age effects among boys born 1930 – 1975. Int. J. Obes. Relat. Metab. Disord. 23, 693 – 701. Varela G, Moreiras O, Carbajal A & Campo M (1995): Estudio nacional de nutricioń y alimentacioń 1991. Madrid: Instituto Nacional de Estadı´stica. Weinsier RL, Hunter GR, Heini AF, Goran MI & Sell SM (1998): The etiology of obesity: relative contribution of metabolic factors, diet, and physical activity. Am. J. Med. 105, 145 – 150.

European Journal of Clinical Nutrition