Introduction

Early interest in leptin centred on its function in regulating energy balance and fat stores, and numerous studies have reported that circulating leptin concentrations are highly correlated with body fat in both adults and children. This regulatory process is still not fully understood. Previous attempts to shed light on it by investigating the relationship between leptin levels and ethnicity have given conflicting results. Leptin levels have been found to be higher in black women compared with white women, of similar body mass index (BMI) and body fat mass,¹ but a number of other studies have failed to find significant differences due to ethnicity.^2,3,4 In this report we compare the leptin–BMI relationship in children from Italy and The Gambia.

Methods

We compared the relationship between gender- and age-specific z-score for BMI and serum leptin concentrations in children from The Gambia, West Africa with that for the Italian children reported previously in this journal.⁵ The Gambian children were drawn from West Kiang, a rural region dependent on subsistence farming. Detailed descriptions of environmental and nutritional conditions in this area can be found elsewhere.^6,7 The children recruited into the study were born during the first 2 y of the West Kiang maternal dietary supplementation study,⁸ and at the time of recruitment were aged between 6 and 10 y. The original Italian study had data from 1929 Italian children aged 3–18 y from three provinces in central Italy (Perugia, Terni, Rieti). Only children aged between 6 and 10 y from the original cohort were used in the analysis presented here. All of the Italian children had a pubertal stage 1, as evaluated according to Tanner and Whitehouse.⁹ Pubertal status was not assessed in the Gambian children. However, previous work in this community has shown that puberty is delayed as compared to British reference children.¹⁰ It is therefore unlikely that children of this age will have reached puberty. All the Italian children in the study were known to be healthy. In the Gambian study, children were not recruited if they were sick and plasma levels of 1-antichymotrypsin were assessed as an indicator of inflammation.

Serum leptin concentrations for the Gambian children were measured at MRC Human Nutrition Research, Cambridge, UK. The Italian samples were analysed at the University of Perugia, Italy.⁵ In both centres, serum leptin concentrations were determined using a well-characterized commercially available human leptin radioimmunoassay kit (Linco Research Inc., St Charles, MO, USA).¹¹ The limit of sensitivity for the Gambian samples was 0.5 ng ml^-1, the intra-assay CV was 5.1%, and the inter-assay CV was 2.8%. For the Italian samples, the intra-assay CV was 6.8% and the inter-assay CV was 7.1%. A cross-validation was performed between sites on a separate set of plasma samples. This analysis (data not presented) demonstrated good correlation between laboratories with a correlation coefficient of 0.983 and a slope not significantly deviating from unity.

Weight and height measurements were measured using standard, regularly validated anthropometric equipment. Full details of the study protocol for each site are provided elsewhere.^5,12 z-scores of BMI for both population groups were calculated using the LMS method of Cole,¹³ based on BMI reference curves for the UK.¹⁴ For the purpose of analysis and in order to normalise the distribution and linearise the relationship with BMI z-score, leptin values were log transformed, and are presented as geometric means. ANOVA was used to test for simple (unadjusted) differences between groups. Analysis of the relationship between leptin and BMI employed multiple regression with linear and quadratic terms, that is, x and (x - )², for BMI z-scores and a linear term for age. The differences between the populations were fitted as intercept effects. The sexes were analysed separately throughout.

Variants of the model including higher-order terms for BMI and age, an interaction between BMI and population ('slope' effect) and weights allow for differences in residual variance between the populations, were also fitted. In order to examine the potential size of the bias induced by random (rather than systematic) variation in the relationship between BMI and body fat (ie for regression towards the mean), we repeated the analysis with an adjusted z-score for BMI. The adjusted score was the weighted mean of the individual's observed, and the population mean, z-score:

where zBMI is the population- and sex-specific mean BMI z-score and r is the correlation coefficient between BMI and % fat mass. Values for r (females: r=0.89; males: r=0.76) were obtained from data on British children.¹⁵ The significance of regression model terms was tested by comparing the reduction in sum of square due to the term with the residual sum of squares using an F-test. The statistical analysis was performed in Stata 7 (Stata Corporation, 4905, Lakeway Drive, College Station, TX 77845, USA).

Results

A total of 471 Gambian children (male=250, females=221) and 839 Italian children (male 435, females 404) were used in the comparison. Details of the two groups are given in Table 1. The Italian children were slightly older (males: Italy 8.6 y, Gambia 8.0 y, P0.0001; females: Italy 8.3 y, Gambia 8.0 y, P=0.0006) and had significantly higher means of BMI z-score (males: Italy 1.58, Gambia –1.44, P0.0001; females: Italy 1.33, Gambia –1.42, P0.0001), and serum leptin concentration (males: Italy 12.45 ng ml^-1, Gambia 1.83 ng ml^-1, P0.0001; females: Italy 15.19 ng ml^-1, Gambia 2.36 ng ml^-1, P0.0001). The gender difference observed in serum leptin concentrations was highly significant in both population groups (0.0001).

Table 1 - Mean (s.d.) age, anthropometry and plasma leptin levels for Gambian and Italian children.

Full table

Figure 1 shows the relationship between serum leptin and BMI z-score, subdivided according to gender. Both Gambian boys and girls had significantly lower levels of leptin (95% CI for boys: 55–65% of Italian values; for girls: 49–59%) than their counterparts of the same age and BMI in Italy (boys: F=143.76 on 1 over 656 d.f., P<0.0001; girls: F=154.23 on 1 over 604 d.f., P<0.0001).

Figure 1.

Serum leptin values vs z-score of BMI, subdivided according to gender. Open circles represent data for Italian children, closed triangles Gambian children. Fitted lines show the least squares, sex- and population-specific quadratic equations.

Full figure and legend (73K)

Table 2 gives the results of fitting the basic regression model for male and female subjects. The quadratic term for age was also significant for male subjects, but not for female subjects, while neither the cubic term for BMI z-score nor the term for the interaction between BMI and country were significant for either sex. The effect of these terms was generally to increase the significance of the difference between countries. Weighting the analysis to allow for a slight difference in residual variance between the two populations also made very little difference to the conclusions.

Table 2 - Results of multiple regression.

Full table

When the BMI z-scores were replaced by their adjusted values, the difference between the two populations fell below significance for boys and was substantially reduced in girls. The adjusted BMI score is even more confounded with country than the unadjusted score: the difference between countries accounts for 70% of its variance.

Comments

There is a striking difference between both the levels of fatness and the leptin levels of these two ecologically diverse populations. The Gambian children had BMI z-scores averaging -1.4 compared to +1.5 for the Italian children; a difference of almost 3 s.d. The ranges also differed greatly (Gambia -5.06 to +0.78 and Italy -2.89 to +4.4 z-scores) with very few Gambian children above the UK reference mean and relatively few Italian children below the mean. The average leptin levels differed by a factor of six-fold, and there was a much greater range in the Italian children (leptin s.d.=0.47 and 0.91 ng ml^-1 in Gambia and 9.94 and 12.25 ng ml^-1 in Italy for males and females, respectively). The slope of leptin against BMI was rather shallow at BMI z-scores below zero and rose steeply at values above zero. We have argued elsewhere¹⁶ that this is likely to have important implications in terms of the position on the dose–response curves regulating leptin's pleiotropic physiological functions at which the hormone is acting. We believe that the very high levels of leptin seen in fatter children will rarely have been encountered in our evolutionary past, and that the Gambian children are probably much more representative of the historical norm. Studies across a wide range of nutritional status, and especially focussing on what are now classified as undernourished children by Western reference standards, will be critical to gaining a full understanding of the evolutionary origins, and hence of the true physiological design characteristics, of the leptin axis.

The results from this analysis also demonstrate that the relationship between BMI and serum leptin levels is considerably different between these two groups of children from very different ethnic backgrounds. At all levels of BMI, the Italian children show a greater level of leptin than the Gambian children. A number of possible explanations for this difference are considered below:

Firstly, it is possible that methodological differences between the two laboratories may have contributed to the differences observed in the leptin levels between the two population groups. However, certain points would indicate that it is unlikely that such differences alone can account for the magnitude of the difference observed between the two populations: All the samples were analysed using the same commercial kit and interassay and intra-assay coefficients of variation for high and low QC samples were low at both centres. A cross-validation of samples exchanged between the two laboratories revealed such close agreement as to exclude methodological bias as a possible source of the differences between the Gambian and Italian children. Furthermore, leptin is a fairly robust analyte, and it is therefore unlikely that differences in sample storage prior to analysis could have led to any differences.

It is also possible that factors known to interfere with the leptin assay, such as the presence of soluble leptin receptors, may differ between population groups. Recent work has indeed demonstrated that gender and body size differences exist in the regulation of circulating soluble leptin receptor levels.¹⁷ It is therefore possible, although no evidence appears to exist it the literature, that ethnic difference in the circulating form of leptin may lead to measurable differences with the standard assay procedures used. Future work would need to explore this area in more detail.

The stage of sexual maturation has been shown to influence leptin levels in adolescents.^18,19 However, the age range of subjects in this study was chosen to ensure all children were prepubertal and therefore avoid this problem. It is also possible that leptin metabolism may be dependent on genetic or environmental factors, which differ between these populations. Leptin is sensitive to the acute state of energy balance as well as to the size of the fat mass.²⁰ A relative state of energy insufficiency in the Gambian subjects and excess in the Italians may contribute to the differences. We note, however, that Luke et al²¹ observed that the relationship between leptin and %fat mass was similar in people of African origin living in Nigeria, Jamaica and the US (with large intergroup differences in BMI).

Alternatively, the differences in the relationship between BMI and leptin might result from a consistent difference in the way in which BMI relates to fat mass between the populations. Such differences can be considerable as Luke et al²¹ reported in their study. Indeed, results from a recent study of children from three different ethnic groups (Singapore Chinese, Beijing Chinese and Dutch) strongly suggest that the relationship between body fat percentage and BMI is different among children of different ethnic background.²² If this is the case, and as Deurenberg et al²² suggest, this observation could have implications for the relevance and use of growth charts and BMI cutoff points for different ethnic groups.

Finally, such differences in the relationship between BMI and leptin may also arise for purely statistical reasons, due to a regression towards the mean artefact. This results from the difference in BMI distribution between the two populations and the fact that BMI is an imperfect measure of body fat. The adjusted score will exactly remove this source of bias provided the error is uniform, that is, that the variance of the mismatch between BMI and % body fat is the same for all values of BMI. This assumption, however, cannot be substantiated. Furthermore, the correlation coefficients between the BMI and % fat mass used to adjust the BMI z-scores, being derived from British children, are possibly not ideal (although in the absence of data from Gambian or Italian children they were the best available). We do not claim, therefore, to have adjusted the BMI scores accurately, only to demonstrate that this problem can introduce considerable bias. In these data, at least for the boys, our adjustment is sufficient to explain the differences in the relationship between leptin and BMI scores for the different populations. These observations have repercussions for all studies of leptin in which BMI is used as a covariable, even when there is less reason to suspect substantial differences in body composition between the populations being compared.

This comparison of two sets of data from very different ethnic groups has revealed an interesting difference in the relationship between BMI and leptin levels. If valid, this finding could have important implications for the use of growth charts and cutoff points for levels of underweight, overweight and obesity in different groups of children. Future studies using detailed measures of energy balance and body composition (dual-energy X-ray absorptiometry, H₂¹⁸O-dilution, etc) to explore the relationship between leptin levels, energy balance and body composition are therefore needed in children from such different ethnic groups.

References

1. der Merwe MT, Panz VR, Crowther NJ, Schlaphoff GP, Gray IP, Froguel P, Joffe BI, Lonnroth PN. Free fatty acids and insulin levels—relationship to leptin levels and body composition in various patient groups from South Africa. Int J Obes Relat Metab Disord 1999; 23: 909–917. | Article | PubMed |
2. Nagy TR, Gower BA, Trowbridge CA, Dezenberg C, Shewchuk RM, Goran MI. Effects of gender, ethnicity, body composition, and fat distribution on serum leptin concentrations in children. J Clin Endocrinol Metab 1997; 82: 2148–2152. | Article | PubMed | ISI | ChemPort |
3. Perry HM, Morley JE, Horowitz M, Kaiser FE, Miller DK, Wittert G. Body composition and age in African-American and Caucasian women: relationship to plasma leptin levels. Metabolism 1997; 12: 1399–1405.
4. Danadian K, Suprasongsin C, Janosky JE, Arslanian S. Leptin in African-American children. J Pediatr Endocrinol Metab 1999; 12: 639–644. | PubMed |
5. Falorni A, Galmacci G, Bini V, Papi F, Molinari D, De Giorgi G, Faraoni F, Celi F, Di Stefano G, Berioli MG, Contessa G, Bacosi ML. Fasting serum leptin levels in the analysis of body mass index cut-off values: are they useful for overweight screening in children and adolescents? A school population-based survey on three provinces of central Italy. Int J Obes Relat Metab Disord 1998; 22: 1197–1208. | Article | PubMed |
6. Prentice AM, Whitehead RG, Roberts SB, Paul AA. Long-term energy balance in childbearing Gambian women. Am J Clin Nutr 1981; 34: 2790–2799. | PubMed | ISI | ChemPort |
7. Prentice AM, Cole TJ, Foord FA, Lamb WH, Whitehead RG. Increased birthweight after prenatal dietary supplementation of rural African women. Am J Clin Nutr 1987; 46: 912–925. | PubMed |
8. Ceesay SM, Prentice AM, Cole TJ, Foord F, Weaver LT, Poskitt EME, Whitehead RG. Effects on birthweight and perinatal mortality of maternal dietary supplementation in a primary health care setting in rural Gambia. Br Med J 1997; 315: 786–790. | ChemPort |
9. Tanner JM, Whitehouse RH. Clinical longitudinal standards for height, weight, height velocity, weight velocity and stages of puberty. Arch Dis Childhood 1976; 51: 170–176.
10. Dibba B, Prentice A, Ceesay M, Stirling DM, Cole TJ, Poskitt EME. Effect of calcium supplementation on bone mineral accretion in Gambian children accustomed to a low-calcium diet. Am J Clin Nutr 2000; 71: 544–549. | PubMed | ISI | ChemPort |
11. Ma Z, Gingerich RL, Satiago JU, Klein S, Smith CH, Landt M. Radioimunassay of leptin in human plasma. Clin Chem 1996; 42: 942–946. | PubMed | ISI | ChemPort |
12. Moore SE, Collinson AC, Prentice AM. Immune function in rural Gambian children is not related to season of birth, birth size, or maternal supplementation status. Am J Clin Nutr 2001; 74: 840–847. | PubMed |
13. Cole TJ, Green PJ. Smoothing reference centile curves: the LMS method and penalized likelihood. Stat Med 1992; 11: 1305–1319. | Article | PubMed | ChemPort |
14. Cole TJ, Freeman JV, Preece MA. Body mass index reference curves for the UK, 1990. Arch Dis Childhood 1995; 73: 25–29. | ISI | ChemPort |
15. Wells JC, Stanley M, Laidlaw AS, Day JM, Davies PS. Energy intake in early infancy and childhood fatness. Int J Obes Relat Metab Disord 1998; 22: 387–392. | Article | PubMed |
16. Prentice AM, Moore SE, Collinson AC, O'Connell MA. Leptin and undernutrition. Nutr Rev 2002; 60: S56–67. | Article | PubMed |
17. Chan JL, Blüher S, Yiannakouris N, Suchard MA, Kratzsch J, Matzoros CS. Regulation of circulating soluble leptin receptor levels by gender, adiposity, sex steroids, and leptin. Observational and intervention studies in humans. Diabetes 2002; 51: 2105–2112. | PubMed | ChemPort |
18. Ellis KJ, Nicolson M. Leptin levels and body fatness in children: effects of gender, ethnicity, and sexual development. Pediatr Res 1997; 42: 484–488. | PubMed |
19. Mantzoros CS, Flier JS, Rogol AD. A longitudinal assessment of hormonal and physical alterations during normal puberty in boys. Rising leptin levels may signal the onset of puberty. J Clin Endocrinol Metab 1997; 82: 1066–1070. | Article | PubMed | ISI | ChemPort |
20. Havel PJ. Role of adipose tissue in body-weight regulation: mechanisms regulating leptin production and energy balance. Proc Nutr Soc 2000; 59: 359–371. | PubMed | ISI | ChemPort |
21. Luke AH, Rotimi CN, Cooper RS, Long AE, Forrester TE, Wilkes R, Bennett FI, Ogunbiyi O, Compton JA, Bowsher RR. Leptin and body composition of Nigerians, Jamaicans, and US blacks. Am J Clin Nutr 1998; 67: 391–396. | PubMed |
22. Deurenberg P, Deurenberg-Yap M, Foo LF, Schmidt G, Wang J. Differences in body composition between Singapore Chinese, Beijing Chinese and Dutch children. Eur J Clin Nutr 2003; 57: 405–409. | Article | PubMed | ISI | ChemPort |

Acknowledgements

We acknowledge Mr Jason Swain, MRC Human Nutrition Research Centre for undertaking the leptin analysis. This work was supported by the Nutricia Research Foundation.