Leptin concentrations in the United States: relations with demographic and anthropometric measures

¹ From Social and Scientific Systems, Inc, Bethesda, MD, and the National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD.

² Supported by a contract from the National Institute of Diabetes and Digestive and Kidney Diseases (NO1-DK-6-2220).

³ Address reprint requests to CE Ruhl, Social and Scientific Systems, Inc, 7101 Wisconsin Avenue, Suite 1300, Bethesda, MD 20814-4805. E-mail: CER{at}s-3.com.

See corresponding editorial on page 277.

ABSTRACT  
Background: Leptin is a peptide that is strongly correlated with adiposity and is a potential determinant of obesity and its complications.

Objective: Leptin concentrations from a representative sample of the US population were examined in relation to demographic and anthropometric measures.

Design: Fasting serum leptin concentrations were measured in 6303 women and men aged 20 y in the third National Health and Nutrition Examination Survey. Anthropometric measures included body mass index, 4 skinfold thicknesses, and 4 body circumferences. Ethnic groups included non-Hispanic whites and blacks and Mexican Americans.

Results: The mean serum leptin concentration was much higher in women (12.7 µg/L) than in men (4.6 µg/L). In a multivariate analysis, leptin concentrations were associated with the sum of 4 skinfold thicknesses, waist and hip circumferences, ethnicity, and age. These measures explained most of the variance in leptin concentrations in women (R² = 0.69) and in men (R² = 0.67). Triceps skinfold thickness, when substituted for the sum of skinfold thicknesses, performed nearly as well in women (R² = 0.68) and men (R² = 0.67). Leptin concentrations were slightly but significantly higher in non-Hispanic blacks than in non-Hispanic whites of both sexes when these anthropometric measures and age were controlled for; Mexican Americans had concentrations that were intermediate compared with the concentrations of non-Hispanic whites and blacks.

Conclusions: In this large, representative sample of the US population, demographic and anthropometric measures predicted serum leptin concentrations in women and men.

Key Words: Leptin • anthropometry • ethnicity • population • epidemiology • obesity • third National Health and Nutrition Examination Survey • NHANES III • United States

INTRODUCTION  
Since its discovery in 1994 (1), leptin, the protein product of the obesity (ob) gene, has generated much attention in the fields of obesity and metabolic research. Leptin is synthesized and secreted by adipocytes, and serum concentrations reflect the amount of energy stored in adipose tissue (2). Leptin binds to receptors in the hypothalamus and influences the expression of several neuropeptides that regulate energy intake, energy expenditure, and neuroendocrine function. Although few persons with extreme obesity are leptin deficient, most obese persons have hyperleptinemia proportionate to body fat and appear to be leptin resistant (2). Although the physiologic role of leptin is incompletely understood, it may be a determinant of obesity and its complications.

Limited attention has been paid to the relation of leptin concentrations with anthropometric measures other than body mass index (BMI) in population-based studies (3–15). Studies conducted in the United States included only subgroups of the population, had limited sample sizes, and used limited numbers of anthropometric measures (12–15). Studies comparing ethnic groups used a limited number of anthropometric measures (13, 15). To increase the understanding of the relation of leptin concentrations with ethnicity and anthropometric measures, we measured serum leptin concentrations in a large, representative, population-based sample from the United States.

SUBJECTS AND METHODS  
Study design
The third National Health and Nutrition Examination Survey (NHANES III) was conducted in the United States from 1988 to 1994 by the National Center for Health Statistics of the Centers for Disease Control and Prevention (CDC). It consisted of interviews, examinations, and laboratory data collected from a complex multistage, stratified, clustered probability sample of the civilian, noninstitutionalized population aged 2 mo, with oversampling of the elderly non-Hispanic blacks and Mexican Americans (16).

Subjects
The study sample consisted of 3366 women and 2937 men aged 20 y who were randomly assigned by household to be examined in the morning at a mobile examination center after an overnight fast. Excluded were persons who were interviewed but not examined (n = 1795) or who were examined at home (n = 457), who were assigned to a morning examination but attended in the afternoon or evening (n = 400), who fasted <9 h (n = 536) or 24 h (n = 9) or whose fasting time was unknown (n = 129), whose serum specimens were unavailable (n = 732) or of insufficient quantity to perform the leptin assay (n = 11), who were pregnant (n = 105), or who had missing BMI data (n = 7). The study was approved by the CDC Institutional Review Board and all participants provided written consent.

Variables
A morning, fasting venous blood sample was collected. Serum specimens were stored at -70°C and went through at least one freeze-thaw cycle during an average of 8 y of storage before leptin concentrations were measured. Leptin was previously shown to remain stable through 5 freeze-thaw cycles (17) and after storage for as long as 29 y (18). Serum leptin concentrations were measured by radioimmunoassay at Linco Research, Inc, St Charles, MO (17). The minimum detectable concentration of the assay is 0.5 µg/L. Intra- and interassay CVs are both <5%.

Height; weight; triceps, subscapular, suprailiac, and thigh skinfold thicknesses; and arm, waist, hip, and thigh circumferences were measured (19). Height was measured by using a stadiometer and weight by using a self-zeroing scale while the subjects were wearing foam slippers and a paper shirt and pants. The triceps skinfold thickness was measured at the midpoint of the right posterior upper arm, the subscapular skinfold thickness directly below and medial to the inferior angle of the right scapula, the suprailiac skinfold thickness at the high point of the right iliac crest in the midaxillary line, and the thigh skinfold thickness at the midpoint of the right anterior thigh. Skinfold thicknesses were measured in duplicate to the nearest 0.1 mm by using skinfold calipers (Holtain Ltd, Crymych, United Kingdom).

Arm circumference was measured at the midpoint of the right upper arm, waist circumference at the high point of the iliac crest, hip circumference at the maximum circumference of the buttocks, and thigh circumference at the midpoint of the right anterior thigh. Circumferences were measured to the nearest 0.1 cm by using a steel measuring tape. BMIs [wt (kg)/ht² (m)], the sum of the 4 skinfold thicknesses, waist-to-hip circumference ratios, and waist-to-thigh circumference ratios were calculated.

Statistical analysis
Because of sex differences in serum leptin concentrations, separate analyses were conducted for women and men. The relation of leptin concentrations with anthropometric measures was examined in a univariate analysis by calculating weighted Pearson's correlation coefficients. Multiple linear regression analysis was used to determine the independent effects of demographic and anthropometric measures on leptin concentrations. In the multivariate analyses, anthropometric variables were entered sequentially based on the proportion of variation in leptin concentrations (R²) that they explained in univariate analysis and were retained if the P value was < 0.10. Quadratic terms for age and anthropometric measures and interaction terms for ethnicity and age and anthropometric measures were also considered for inclusion by adding them individually to the multivariate models. Analysis of residuals was performed to evaluate the appropriateness of the final multivariate models. Because leptin concentrations were skewed to the right, log₁₀ transformation was performed to normalize the distribution for calculation of correlation coefficients and for regression analyses; the result was back-transformed to produce geometric means. Multivariate analyses excluded persons with missing values for any factor included in the model. A 95% CI that did not include 1 (P < 0.05) was considered to indicate statistical significance. The analyses incorporated sample weights, stratification, and clustering by using SUDAAN software (20).

RESULTS  
Fasting serum leptin concentrations ranged from the minimal detectable concentration of 0.5 to 192.5 µg/L. Women of all ethnic groups had substantially higher fasting serum leptin concentrations than did men (Table 1). The median leptin concentration in women was slightly lower than the 95th percentile in men. In women, the leptin concentration was highest in non-Hispanic blacks followed by Mexican Americans and non-Hispanic whites (P < 0.01 for all comparisons). In men, the leptin concentration was significantly higher in non-Hispanic whites than in non-Hispanic blacks (P = 0.01) and was higher in non-Hispanic whites than in Mexican Americans, but not significantly so. Leptin concentrations increased with age (Appendix A) and markedly increased as BMI increased (Appendix B) in both women and men (P < 0.001 for all comparisons).

View this table:
TABLE 1. Serum leptin concentrations in non-Hispanic white, non-Hispanic black, and Mexican American women and men aged 20 y: United States, 1988–1994  

View this table:
APPENDIX A . Serum leptin concentrations in women and men aged 20 y by ethnicity and age: United States, 1988–1994¹  

View this table:
APPENDIX B . Serum leptin concentrations in women and men aged 20 y by ethnicity and BMI: United States, 1988–1994¹  
The ethnic distribution and anthropometric measures for women and men are shown in Table 2. Leptin concentrations were strongly correlated with various anthropometric measures in women and men, both in the univariate analysis and when age and ethnicity were controlled for (Table 3). Correlations between leptin concentrations and skinfold thicknesses diminished after age, ethnicity, and BMI were controlled for, but were still highly significant. Correlations of leptin with body circumferences and circumference ratios diminished further after BMI was controlled for. Conversely, the correlation with BMI declined, but was still significant after age, ethnicity, and the sum of skinfold thicknesses were controlled for. After control for BMI and the sum of skinfold thicknesses, the correlation of BMI with leptin was diminished more so than was the correlation of the sum of skinfold thicknesses with leptin.

View this table:
TABLE 2. Characteristics of the women and men in the study sample  

View this table:
TABLE 3. Correlation of serum leptin concentration with anthropometric measures in women and men¹  
In women, waist and hip circumferences and the sum of skinfold thicknesses were the anthropometric variables that most strongly predicted leptin concentrations in the multivariate linear regression analysis (Table 4). There was a slight positive association of leptin with non-Hispanic black ethnicity after the sum of skinfold thicknesses, waist and hip circumferences, and age were controlled for. Mexican American women had higher leptin concentrations than did non-Hispanic whites, but the difference appeared to decline as the sum of skinfold thicknesses increased (ie, there was a negative interaction between Mexican American ethnicity and the sum of skinfold thicknesses). This relation is depicted in Figure 1, which shows actual leptin concentrations for the 3 major ethnic groups according to the sum of skinfold thicknesses. Leptin concentrations increased in all groups as the sum of skinfold thicknesses increased, but somewhat more slowly in Mexican American women in the 50th through the 90th percentiles.

View this table:
TABLE 4. Demographic and anthropometric associations with log₁₀ serum leptin concentrations (µg/L) in women and men  

View larger version (15K):
FIGURE 1. . Geometric mean (±SE) serum leptin concentrations in women by percentile for the sum of 4 skinfold thicknesses and ethnicity. Decile cutoffs (mm) for the 10th to 90th percentiles, respectively: <51.8, 51.8 to <63.3, 63.3 to <73.9, 73.9 to <83.9, 83.9 to <93.7, 93.7 to <104.3, 104.3 to <116.4, 116.4 to <128.9, 128.9 to <148.8 and 148.8.

 
In men, the sum of skinfold thicknesses and waist circumference predicted leptin concentrations in the multivariate analysis. There was a slight positive association of leptin concentrations with non-Hispanic black ethnicity and a nonsignificantly higher concentration in Mexican Americans after the sum of skinfold thicknesses, waist circumference, and age were controlled for. Actual mean leptin concentrations by the sum of skinfold thicknesses are shown for the 3 ethnic groups in Figure 2.

View larger version (14K):
FIGURE 2. . Geometric mean (±SE) serum leptin concentrations in men by percentile for the sum of 4 skinfold thicknesses and ethnicity. Decile cutoffs (mm) for the 10th to 90th percentiles, respectively: <36.7, 36.7 to <45.3, 45.3 to <53.4, 53.4 to <59.7, 59.7 to <65.5, 65.5 to <72.3, 72.3 to <79.1, 79.1 to <87.8, 87.8 to <102.5 and 102.5.

 
Of note, BMI did not predict leptin concentrations in the multivariate analysis for either women or men. Because BMI was correlated with the sum of skinfold thicknesses and body circumferences, the lack of an association of leptin concentrations with BMI, independent of these measures, was verified by regressing the residuals of leptin concentrations from the final multivariate models on BMI. BMI was not significantly associated with the residuals in women or men, indicating that BMI made no independent contribution to the final multivariate models.

The sum of skinfold thicknesses was the major determinant of leptin concentrations. Of the skinfold thicknesses, triceps skinfold thickness had the highest partial correlation coefficient when substituted for the sum of skinfold thicknesses in the multivariate model (data not shown). To simplify the measurements needed to calculate predicted leptin concentrations, we substituted the triceps-skinfold-thickness measurement for the sum of skinfold thicknesses in the multivariate model. In women, this substitution resulted in a decline in the overall R² value from 0.69 to 0.68; all other variables in the model remained significant, including the interaction between Mexican American ethnicity and triceps skinfold thickness (Table 5). In men, the overall R² value remained 0.67 after substitution of the triceps skinfold thickness for the sum of skinfold thicknesses in the model. Unlike in the model for women, age and waist circumference squared were no longer significantly associated with leptin concentrations after substitution of triceps skinfold thickness for the sum of skinfold thickness in the model for men. Therefore, only ethnicity, triceps skinfold thickness, and waist circumference were significant predictors of leptin concentrations in men.

View this table:
TABLE 5. Demographic and anthropometric associations with log₁₀ serum leptin concentrations (µg/L) in women and men when triceps skinfold thickness was substituted in the model for the sum of 4 skinfold thicknesses  

DISCUSSION  
In this large, representative sample of the US population, leptin concentrations were strongly associated with female sex, as was consistently found in previous studies. Non-Hispanic blacks and Mexican Americans had slightly higher leptin concentrations than did non-Hispanic whites when anthropometric variables were controlled for. However, these differences were small, with predicted concentrations that were 6–8% higher in non-Hispanic black women and men than in non-Hispanic white women and men. Concentrations were also higher in Mexican American women than in non-Hispanic white women, but were predicted to vary from 15% higher at the 5th percentile of the sum of skinfold thicknesses (45 cm) to 5% higher at the 50th percentile (94 cm) and to 8% lower at the 95th percentile (161 cm). It is uncertain whether any physiologic significance can be ascribed to such relatively small differences across ethnic groups, particularly when measures of body fat are imprecise.

Few previous population-based studies examined leptin concentrations in relation to ethnicity, and all these studies had a smaller sample size than did the present study. A study from Chile found lower leptin concentrations in Mapuche natives than in Hispanics (21), but most other studies found no major ethnic differences. Similarly to the findings in the present study, Mexican Americans had slight but nonsignificantly higher leptin concentrations than did non-Hispanic whites after adjustment for BMI in a study in San Antonio, TX (22). Leptin concentrations did not differ significantly between African Americans, Cuban Americans, and non-Hispanic whites in Dade County, FL, in a multivariate analysis in which the percentage of body fat was adjusted for (15). The findings were inconsistent in non-population-based studies of leptin concentrations and ethnicity (23–25).

The prevalence of overweight and obesity varies among ethnic groups, particularly among women. In the women in NHANES III, the prevalence of a BMI 27.3 was 34% in non-Hispanic whites, 52% in non-Hispanic blacks, and 50% in Mexican Americans (26). Although the present study was limited by the inability of anthropometric measures such as skinfold thicknesses and body circumferences to fully adjust for differences in body composition among ethnic groups, our findings provide further evidence that ethnicity is of limited import in determining serum leptin concentrations in these 3 groups. The small differences in leptin concentrations between the 3 ethnic groups do not suggest that leptin can explain the observed marked disparities in overweight between the 3 groups.

Predicted leptin concentrations can be obtained by multiplying the antilog₁₀ of the sum of the values for the measured variables from Table 4 or 5 by their regression coefficients. Fewer measures of skinfold thickness are provided in Table 5 than in Table 4—1 skinfold-thickness measure instead of 4—resulting in a small sacrifice in precision. For example, the predicted leptin concentration in non-Hispanic black men from the data in Table 5 would be as follows: antilog₁₀[-0.93 + 0.045 + 0.011(value for waist circumference, in cm) + 0.040(value for triceps skinfold thickness, in mm) - 0.00062(square of the value for triceps skinfold thickness) + 0.0020(value for age, in y)]. These prediction equations are most useful in clinical studies to determine whether concentrations in a particular group of individuals are higher or lower than those predicted from a national reference population.

Skinfold thicknesses dominated the prediction equations for leptin concentrations. This finding suggests that measures of peripheral and subcutaneous fat are a stronger determinant of leptin concentrations than are more global measures such as BMI, which does not discriminate between fat and lean body mass or between specific fat depots. This conclusion is supported by the finding of an association of leptin concentrations with total and subcutaneous fat, but not with visceral fat (27–29), and a greater production of leptin in subcutaneous than in visceral fat (30, 31).

The relation of leptin concentrations with measures of body fat distribution was examined in other population studies. A study in Mexican Americans also found high correlations of leptin concentrations with BMI and waist and hip circumferences and concluded that leptin concentrations are associated with overall adiposity rather than with a specific fat depot (12). In contrast, several authors found an association of leptin concentrations with waist circumference, independent of BMI or percentage body fat, and concluded that body fat distribution may also be an important determinant of leptin concentrations (3, 4, 8). In another study, leptin concentrations were unrelated to waist circumference after adjustment for fat mass, but were associated with hip circumference in women (6). In our study, leptin concentrations were associated with both waist and hip circumferences and with skinfold thicknesses, independent of BMI. Rather than indicating a role for specific fat depots in determining leptin concentration, this finding may be the result of limitations of BMI as a surrogate for total fat mass.

Although anthropometric measures are less accurate than are measures of total fat mass and specific fat depots by methods such as computed tomography or dual-energy X-ray absorptiometry, we had no information from such methods. However, our final multivariate models, which used subcutaneous fat (represented by the sum of skinfold thicknesses) and visceral fat (reflected by waist circumference), explained a degree of variance in leptin concentrations at least as high as that explained by models containing total fat mass (R² = 49–65%) or percentage body fat (R² = 52–65%) measured by computed tomography, dual-energy X-ray absorptiometry, or bioelectrical impedance analysis (5, 7, 13, 32).

NHANES III is unique in that it provides national, population-based, multiethnic measures. Examination of leptin concentrations in the present population was facilitated by the carefully performed and well-described anthropometric measurements (19). Simple sets of similar anthropometric and demographic measures explained most of the variance in serum leptin concentrations in both women and men. The associations of leptin concentrations with demographic and anthropometric measures should prove useful in relation to other measures in NHANES III and as national reference measures.

APPENDIX A  

APPENDIX B  

ACKNOWLEDGMENTS  
We thank Keith Rust for statistical advice, Danita Byrd-Holt for assisting with computer programming, and the National Center for Health Statistics staff for making stored serum available and for reviewing the study design.

REFERENCES  

1. Zhang Y, Proenca R, Maffei M, Barone M, Leopold L, Friedman JM. Positional cloning of the mouse obese gene and its human homologue. Nature 1994;372:425–32.
2. Considine RV, Sinha MK, Heiman ML, et al. Serum immunoreactive-leptin concentrations in normal-weight and obese humans. N Engl J Med 1996;334:292–5.
3. Zimmet P, Hodge A, Nicolson M, et al. Serum leptin concentration, obesity, and insulin resistance in Western Samoans: cross sectional study. BMJ 1996;313:965–9.
4. Hanley AJG, Harris SB, Gao XJ, Kwan J, Zinman B. Serum immunoreactive leptin concentrations in a Canadian aboriginal population with high rates of NIDDM. Diabetes Care 1997;20:1408–15.
5. Ramachandran A, Snehalatha C, Vijay V, Satyavani K, Latha E, Haffner SM. Plasma leptin in non-diabetic Asian Indians: association with abdominal adiposity. Diabet Med 1997;14:937–41.
6. Bennett FI, McFarlane-Anderson N, Wilks R, Luke A, Cooper RS, Forrester TE. Leptin concentration in women is influenced by regional distribution of adipose tissue. Am J Clin Nutr 1997;66:1340–4.
7. Solin MS, Ball MJ, Robertson I, et al. Relationship of serum leptin to total and truncal body fat. Clin Sci 1997;93:581–4.
8. de Silva A, de Courten M, Zimmet P, et al. Lifestyle factors fail to explain the variation in plasma leptin concentrations in women. Nutrition 1998;14:653–7.
9. Hodge AM, de Courten MP, Dowse GK, et al. Do leptin levels predict weight gain? A 5-year follow-up study in Mauritius. Obes Res 1998;6:319–25.
10. Ruige JB, Dekker JM, Blum WF, et al. Leptin and variables of body adiposity, energy balance, and insulin resistance in a population-based study. The Hoorn Study. Diabetes Care 1999;22:1097–104.
11. Neuhauser-Berthold M, Herbert BM, Luhrmann PM, et al. Resting metabolic rate, body composition, and serum leptin concentrations in a free-living elderly population. Eur J Endocrinol 2000;142:486–92.
12. Haffner SM, Gingerich RL, Miettinen H, Stern MP. Leptin concentrations in relation to overall adiposity and regional body fat distribution in Mexican Americans. Int J Obes Relat Metab Disord 1996;20:904–8.
13. Luke AH, Rotimi CN, Cooper RS, et al. Leptin and body composition of Nigerians, Jamaicans, and US blacks. Am J Clin Nutr 1998; 67:391–6.
14. Folsom AR, Jensen MD, Jacobs DR Jr, et al. Serum leptin and weight gain over 8 years in African American and Caucasian young adults. Obes Res 1999;7:1–8.
15. Donahue RP, Zimmet P, Bean JA, et al. Cigarette smoking, alcohol use, and physical activity in relation to serum leptin levels in a multiethnic population: The Miami Community Health Study. Ann Epidemiol 1999;9:108–13.
16. National Center for Health Statistics, Centers for Disease Control. Plan and operation of the Third National Health and Nutrition Examination Survey, 1988–94. Washington, DC: US Government Printing Office, 1994. [Series 1, 32. DHHS publication (PHS) 94-1308.]
17. Ma Z, Gingerich RL, Santiago JV, Klein S, Smith CH, Landt M. Radioimmunoassay of leptin in human plasma. Clin Chem 1996;42: 942–6.
18. Lissner L, Karlsson C, Lindroos AK, et al. Birth weight, adulthood BMI, and subsequent weight gain in relation to leptin levels in Swedish women. Obes Res 1999;7:150–4.
19. NCHS. National Health and Nutrition Examination Survey III: body measurements (anthropometry). Rockville, MD: Westat, Inc, 1988 (CD-ROM).
20. Research Triangle Institute. SUDAAN user's manual release 7.0. Research Triangle Park, NC: Research Triangle Institute, 1996.
21. Pérez-Bravo F, Albala C, Santos JL, et al. Leptin levels distribution and ethnic background in two populations from Chile: Caucasian and Mapuche groups. Int J Obes Relat Metab Disord 1998;22:943–8.
22. Wei M, Stern MP, Haffner SM. Serum leptin levels in Mexican Americans and non-Hispanic whites: association with body mass index and cigarette smoking. Ann Epidemiol 1997;7:81–6.
23. Nicklas BJ, Toth MJ, Goldberg AP, et al. Racial differences in plasma leptin concentrations in obese postmenopausal women. J Clin Endocrinol Metab 1997;82:315–7.
24. Perry HM III, Morley JE, Horowitz M, et al. Body composition and age in African-American and Caucasian women: relationship to plasma leptin levels. Metabolism 1997;46:1399–405.
25. Mills PJ, Ziegler MG, Morrison TA. Leptin is related to epinephrine levels but not reproductive hormone levels in cycling African-American and Caucasian women. Life Sci 1998;63:617–23.
26. Anonymous. Update: prevalence of overweight among children, adolescents, and adults—United States, 1988–1994. MMWR Morb Mortal Wkly Rep 1997;46:198–202.
27. Dua A, Hennes MI, Hoffmann RG, et al. Leptin: a significant indicator of total body fat but not of visceral fat and insulin insensitivity in African-American women. Diabetes 1996;45:1635–7.
28. Takahashi M, Funahashi T, Shimomura I, Miyaoka K, Matsuzawa Y. Plasma leptin levels and body fat distribution. Horm Metab Res 1996;28:751–2.
29. Langendonk JG, Pijl H, Toornvliet AC, et al. Circadian rhythm of plasma leptin levels in upper and lower body obese women: influence of body fat distribution and weight loss. J Clin Endocrinol Metab 1998;83:1706–12.
30. Hube F, Lietz U, Igel M, et al. Difference in leptin mRNA levels between omental and subcutaneous abdominal adipose tissue from obese humans. Horm Metab Res 1996;28:690–3.
31. Montague CT, Prins JB, Sanders L, Digby JE, O'Rahilly S. Depot- and sex-specific differences in human leptin mRNA expression: implications for the control of regional fat distribution. Diabetes 1997;46:342–7.
32. Echwald SM, Clausen JO, Hansen T, et al. Analysis of the relationship between fasting serum leptin levels and estimates of beta-cell function and insulin sensitivity in a population sample of 380 healthy young Caucasians. Eur J Endocrinol 1999;140:180–5.