Comparison of Self-reported and Measured BMI as Correlates of Disease Markers in U.S. Adults


  • Mara A. McAdams,

    1. Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts
    Search for more papers by this author
  • Rob M. Van Dam,

    1. Department of Nutrition, Harvard School of Public Health, Boston, Massachusetts
    Search for more papers by this author
  • Frank B. Hu

    Corresponding author
    1. Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts
    2. Department of Nutrition, Harvard School of Public Health, Boston, Massachusetts
    3. Channing Laboratory, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts.
    Search for more papers by this author

Department of Nutrition, Harvard School of Public Health, 665 Huntington Avenue, Boston, MA 02115. E-mail:


Objective: The purpose of this study is to evaluate the validity of BMI based on self-reported data by comparison with technician-measured BMI and biomarkers of adiposity.

Research Methods and Procedures: We analyzed data from 10,639 National Health and Nutrition Education Study III participants ≥20 years of age to compare BMI calculated from self-reported weight and height with BMI from technician-measured values and body fatness estimated from bioelectrical impedance analysis in relation to systolic blood pressure, fasting blood levels of glucose, high-density lipoprotein-cholesterol, triglycerides, C-reactive protein, and leptin.

Results: BMI based on self-reported data (25.07 kg/m2) was lower than BMI based on technician measurements (25.52 kg/m2) because of underreporting weight (−0.56 kg; 95% confidence interval, −0.71, −0.41) and overreporting height (0.76 cm; 95% confidence interval, 0.64, 0.88). However, the correlations between self-reported and measured BMI values were very high (0.95 for whites, 0.93 for blacks, and 0.90 for Mexican Americans). In terms of biomarkers, self-reported and measured BMI values were equally correlated with fasting blood glucose (r = 0.43), high-density lipoprotein-cholesterol (r = −0.53), and systolic blood pressure (r = 0.54). Similar correlations were observed for both measures of BMI with plasma concentrations of triglycerides and leptin. These correlations did not differ appreciably by age, sex, ethnicity, or obesity status. Correlations for percentage body fat estimated through bioelectrical impedance analysis with these biomarkers were similar to those for BMI.

Discussion: The accuracy of self-reported BMI is sufficient for epidemiological studies using disease biomarkers, although inappropriate for precise measures of obesity prevalence.


Many large epidemiological studies depend on self-reported weight and height as surrogates for technician measurements (1, 2). Self-reported measures are more feasible to collect for large population samples, less burdensome for the study participants, and entail fewer costs than values obtained by clinical measurements of height and weight. Although there are many practical advantages of using self-reported data, adiposity values calculated from these measures, such as BMI, may be subject to inaccuracies because of random and systematic errors, which in turn affect the validity of the measures (3). For this reason, it is important to quantify the degree to which self-reported values agree with clinical measurements and whether these participant-provided responses are sufficiently accurate to be used in epidemiological studies.

More than two decades ago, Stunkard and Albaum (4) reported a high accuracy of self-reported weights across different ages and sexes. Also, Stewart (3) studied more than 3000 subjects 14 to 61 years of age and found that self-reported values are valid and reliable indicators of measured weight and height, a finding that was confirmed by subsequent studies (5, 6, 7, 8, 9, 10, 11). This work has expanded from weight and height to BMI and to racial/ethnic subgroups through larger, nationally representative U.S. studies (6, 10). Data from the Health and Nutrition Education Study III (NHANES III)1 has shown that there is an increasing prevalence of obesity in the United States (12), and this dataset has been used to assess the validity of classification of overweight and obesity through the use of self-reported weight and height by ethnicity, sex, and age (6, 7, 10). Gillum et al. (10) found a substantial underestimation of the prevalence of obesity based on self-reported height and weight in women and Mexican-American men.

We have found no studies that compared the correlations of biomarkers of disease with BMI based on self-reports and technician measurements. While traditional anthropometric measures such as weight and BMI are often used in epidemiological studies, measures such as bioelectrical impedance analysis (BIA) are used with increasing frequency to better distinguish between fat and lean tissue (13, 14). The aim of this study is to assess the validity of BMI based on self-reports in a nationally representative sample of U.S. adults (NHANES III) by 1) comparison with BMI based on technician measurements and 2) comparison of correlations for BMI based on self-reports, BMI based on technician measurements, and percent body fat (PBF) based on BIA with disease biomarkers.

Research Methods and Procedures


The NHANES III study was conducted between 1988 and 1994 by the National Center for Health Statistics of the Centers for Disease Control and Prevention. This study was designed to provide a nationally representative sample of the U.S. population (≥2 months of age) and to collect information on the health status of U.S. civilian noninstitutionalized populations. NHANES III sample design reflects a stratified multistage probability design that included oversampling of Mexican-American and black populations. This was done to produce statistically reliable health estimates for the two largest ethnic minority groups in the United States. The sampling method was similar to those used in previous Health and Nutrition Surveys, and details of the sampling methods and methodological information have been previously published (15, 16).

NHANES III consisted of a survey and clinical examination. All participants were interviewed at home (N = 39,695), with 85.6% (n = 33,994) invited to participate in the clinical examination portion; 31,311 of those invited received a clinical examination. The exam was performed in a mobile examination center for all who were able and at home for those who were unable to attend the mobile examination center.


Weight and height were measured by two methods: self-reported and clinically examined. Self-reported measures included in the at home survey, and clinical measurements were collected at the clinical examination. Respondents were asked, “How much do you weigh without clothes and shoes,” and “How tall are you without shoes” (15). These self-reported values were recorded in inch and pound units and converted to metric units. Subsequently, trained health technicians measured the weight and height of those participants who attended the clinical examination, as described previously (15). Height was measured to the nearest millimeter using a fixed stadiometer. Body weight was measured in kilograms (to the second decimal place) by a self-zeroing digital weight scale for adults dressed in underpants, a disposable paper gown, paper pants, and foam slippers. A Toledo 2181 self-zeroing digital weight scale (Toledo Scale, Columbus, OH) was used at the mobile examination center, and a SECA Integra Model 815 Scale (SECA, Rumily, France) was used for the home examination. Both scales were standardized (15). Measured values refer to weight and height measured by a physician during the clinical examination.

For the first time in a NHANES study, bioelectrical impedance resistance measures were also taken during the clinical examination. To obtain PBF through BIA, patients were instructed to lie on their backs while electrodes were placed on their wrists and ankles. These electrodes delivered a low-level alternating current (<1 mA) measured at distinct frequencies between 5 KHz and 1 MHz (15). PBF from BIA was used as a measure of body fat composition that may not be captured through BMI measures. PBF was calculated using prediction equations (14). First fat-free mass (FFM) was calculated through the Deurenberg prediction equation (17):


(where M = 1, F = 0)

Total body fat (TBF) was defined as the difference between weight and FFM. Finally, PBF was calculated as the ratio of TBF to weight (kg).


Those participating in the survey portion of NHANES III were not told that they might be weighed and measured later during the clinical examination. Blinding the participants to the fact that they would later be clinically measured helps ensure that the self-reported measures will be more similar to self-reported weight and height collected in epidemiological studies. The time lapse between the survey and clinical measurements at the mobile examination center was 4 weeks (15).

Continuous measures of BMI were calculated by dividing weight in kilograms by height in meters squared. Measured BMI was calculated using the weight and height recorded in the physician's examination portion of the survey. The term measured BMI is used here to describe BMI calculated using measured weight and height. Self-reported BMI was calculated from the previously mentioned questions, which were part of the household survey; self-reported BMI will be used to describe BMI calculated by self-reported weight and height.

Fasting blood levels of glucose (FBG), triglycerides (TGs), blood levels of high-density lipoprotein-cholesterol (HDLC), systolic blood pressure (SBP), C-reactive protein (CRP), and leptin (LEP) were measured in blood collected during the clinical examination (18).

Analytic Sample

All subjects in this analysis were participants in NHANES III. Of the 31,311 individuals ≥2 months of age who participated in the clinical examination, 10,639 met the inclusion criteria. For this analysis, we first excluded all participants who were <20 years old (14,281 participants). Of the adult population, we also excluded those with reported diabetes or diabetes treatment (1437 participants), and, in addition, any of the remaining participants who used antihypertensive medication or cholesterol-lowering treatments (4162 participants). These treatments and diseases could confound the examined associations and complicate interpretation of the results. Finally, participants who were missing self-reported or measured height or weight (792 participants) were excluded from this analysis, so that each patient would have at least one self-reported measure of BMI and one anthropometric measure. In addition, women who were pregnant and participants with pacemakers were excluded to ensure the accuracy of the BIA resistance data. Sample sizes varied for the biological markers because of missing values, with all analyses containing >10,000 participants except for the analysis involving LEP (n = 4203). After all exclusions, 5239 men and 5400 women were included in the full analysis; 4940 men and 5080 women in the analysis for FBG; 4952 men and 5087 women for TG; 4929 men and 5078 women for HDL; 4973 men and 5122 women for SBP; 4932 men and 5074 women for CRP; and 2007 men and 2196 women for LEP. Of the participants >12 years of age, 17,726 patients had BIA resistance measures, of which we analyzed 9481 who met our inclusion criteria.

Statistical Analysis

This study analyzed NHANES III data with SAS (version 8; Statistical Analysis System, Cary, NC) and SUDAAN (Research Triangle Institute, Research Triangle Park, NC) software (19) to incorporate the sampling weights consistent with the sample design of NHANES III. This was done to compensate for the unequal probability of selection caused by the complex, multistage design of the survey, adjustment for noncoverage and nonresponse, and oversampling of subgroups. SUDAAN was used to calculate estimates and SE by taking into account the sampling weights and sample design. Statistical sampling weights for the total mobile examination center or the home clinically examined samples were used, when appropriate.

The error in self-reported weight was calculated as the difference between self-reported and measured weight. Differences were also calculated for height and BMI, for which we calculated 95% confidence intervals (CIs). Error in self-reported weight, height, and BMI was calculated for groups differing in sex, race/ethnicity, and age.

Adults were categorized by sex (male or female), age group (<60 or ≥60 years of age at interview), and BMI category (<25, 25 to 29.9, or ≥30 kg/m2). Means and SE were calculated for self-reported and measured weight, height, and BMI, as well as biological measures. Age was categorized as <60 or ≥60 years to be consistent with previous published papers that suggested an effect of age on the validity of self-reported weight, height, and BMI (12). BMI was categorized using the established World Health Organization cut-points (20). Tables were stratified by age, race/ethnicity, and sex.

To further study the association between self-reported and measured BMI, we constructed a Bland-Altman plot (21) of the differences in measured and self-reported BMI vs. the average of these two values.

Outcome variables with skewed distributions, such as self-reported weight and height, measured weight and height, FBG, TG, HDL, SBP, CRP, and LEP were log-transformed for the analyses. Adjusted multiple Pearson's correlations were used to measure the correlation between log-transformed self-reported variables and log-transformed measured variables, as well as log-transformed biological measures adjusted for potential confounders. These adjustments took into account age, sex, and race/ethnicity variation in the previously listed categories. In the stratified analyses, we adjusted for all confounders excluding the stratification factor. We chose to use estimates that were adjusted for potential confounders to evaluate the validity of self-reported BMI because key sociodemographic variables such as age, sex, and ethnicity are typically adjusted in an epidemiological study of adiposity measures and health outcomes. Also included in these models were categorical alcohol consumption in the past year (four groups dependent on the quartiles of frequency and quantity) and current smoking status (current smoker or current nonsmoker). All presented p values are two-sided.


Population Characteristics

Table 1 shows the summary statistics for the anthropometric measurements and biological markers of disease by sex, race/ethnicity, and age group. Compared with men, women had lower FBG and SBP but higher HDL and LEP. Most notably, TG was higher in the Mexican-American men than other groups and higher for participants ≥60 years of age than for younger participants. Similar summary statistics were found among subgroups by sex, race/ethnicity, and age (Table 1).

Table 1.  Descriptive statistics for biomarkers in NHANES III
  1. Values are means (SE).

Measured weight (kg)10,63973.16 (0.32)80.15 (0.39)66.38 (0.40)81.11 (0.45)65.73 (0.46)80.86 (0.37)72.96 (0.58)
Measured height (m)10,6391.69 (0)1.76 (0)1.62 (0)1.77 (0)1.63 (0)1.77 (0)1.63 (0)
Measured BMI (kg/m2)10,63925.52 (0.10)25.85 (0.10)25.19 (0.15)25.89 (0.13)24.78 (0.17)25.86 (0.11)27.30 (0.23)
FBG (mg/dL)10,02093.07 (0.25)95.35 (0.28)90.82 (0.35)95.05 (0.32)90.63 (0.44)94.21 (0.61)90.03 (0.45)
HDL (mg/dL)10,00751.39 (0.35)46.27 (0.40)56.41 (0.44)45.49 (0.51)56.78 (0.55)53.21 (0.55)58.00 (0.57)
SBP (mmHg)10,095116.90 (0.36)120.37 (0.43)113.52 (0.43)120.47 (0.48)113.66 (0.52)122.14 (0.53)114.67 (0.48)
CRP (mg/dL)10,0060.37 (0.01)0.34 (0.01)0.39 (0.01)0.34 (0.01)0.37 (0.02)0.37 (0.01)0.50 (0.02)
TG (mg/dL)10,039128.48 (1.94)143.96 (2.98)113.26 (1.82)143.96 (3.42)114.35 (2.35)114.68 (2.17)94.20 (1.73)
LEP (F/liter)4,20310.28 (0.25)5.29 (0.17)14.96 (0.41)5.33 (0.22)14.33 (0.46)4.93 (0.19)18.21 (0.54)

Comparison of BMI Based on Self-reports and BMI Based on Technician Measurements

The Pearson's correlation coefficients between self-reported and measured weight, by sex, race/ethnicity, and age group, are presented in Table 2. All correlation coefficients describing the relationship between self-reported and measured weight were 0.92 or greater. There was greater variation in the correlations for height than for weight by sex, race/ethnicity, and age, with the lowest correlations for Mexican-American men <60 years (r = 0.83) and Mexican-American women >60 years (r = 0.73). BMI calculated by self-reported variables was highly correlated with that calculated by measured variables in the white (r = 0.95), black (r = 0.93), and Mexican-American (r = 0.90) populations. Among the subgroups by age and sex, all correlation coefficients for measured and self-reported BMI were between 0.87 and 0.95.

Table 2.  Correlation between self-reported and examined weight, height, and BMI by sex, age, and ethnicity
 WhiteBlackMexican- American
 20 to 59 years   
 60+ years   
 20 to 59 years   
 60+ years   

Table 3 presents differences between measured and self-reported weight and height. BMI based on self-reported data (25.07 kg/m2) was lower than BMI based on technician measurements (25.52 kg/m2) because of underreporting weight (−0.56 kg; 95% CI, −0.71, −0.41) and overreporting height (0.76 cm; 95% CI, 0.64, 0.88). Women underreported weight on average of −1.47 kg (95% CI, −1.66, −1.27), and the group of women with the highest BMI (−4.48 kg; 95% CI, −5.13, −3.84) had the largest differences in self-reported and measured weight. Height was overreported by varying amounts in all subgroups. On average, men overreported their height to a greater degree than women (1.16 vs. 0.37 cm). Differences in the two measures of height were greater in the highest BMI group and in those >60 years (2.65 cm; 95% CI, 2.48, 2.83). The largest error in BMI occurred in Mexican-American women whose measured BMI was >30 kg/m2 (−2.25 kg/m2; 95% CI, −2.56, −1.94).

Table 3.  Differences and SD in measured and self-reported weight (kg), height (cm), and BMI (kg/m2) by measured BMI
    BMI (<25)BMI (25 to 29.9)BMI (≥30)
  • *

    Statistically not significant.

Total−0.56 (0.08)0.76 (0.06)−0.45 (0.03)0.45 (0.07)0.47 (0.08)0.03 (0.03)*−0.75 (0.09)0.97 (0.08)−0.57 (0.04)−3.54 (0.24)1.30 (0.12)−1.77 (0.09)
 Men0.37 (0.11)1.16 (0.08)−0.22 (0.04)1.52 (0.11)0.98 (0.11)0.25 (0.04)−0.02 (0.13)*1.19 (0.10)−0.36 (0.05)−2.33 (0.35)1.68 (0.19)−1.36 (0.13)
 Women−1.47 (0.10)0.37 (0.06)−0.67 (0.04)−0.37 (0.07)0.07 (0.08)*−0.15 (0.03)−1.90 (0.12)0.63 (0.11)−0.89 (0.06)−4.48 (0.32)1.00 (0.15)−2.09 (0.14)
 Men0.25 (0.13)*1.32 (0.09)−0.31 (0.05)1.34 (0.13)1.16 (0.13)0.14 (0.05)−0.07 (0.16)*1.31 (0.12)−0.42 (0.06)−2.34 (0.45)1.89 (0.24)−1.44 (0.17)
 Women−1.45 (0.12)0.20 (0.08)−0.61 (0.05)−0.46 (0.09)0.09 (0.09)*−0.14 (0.04)−2.07 (0.16)0.49 (0.13)−0.92 (0.08)−4.43 (0.48)0.91 (0.19)−2.05 (0.20)
 Men1.08 (0.15)1.07 (0.09)0.05 (0.05)*3.03 (0.16)0.80 (0.14)0.79 (0.05)0.25 (0.16)*1.19 (0.13)−0.27 (0.06)−2.90 (0.50)1.61 (0.21)−1.47 (0.18)
 Women−1.86 (0.17)0.37 (0.10)−0.79 (0.08)0.13 (0.09)*0.17 (0.10)*0.03 (0.05)*−1.84 (0.22)0.35 (0.19)−0.75 (0.11)−5.00 (0.52)0.72 (0.21)−2.12 (0.23)
 Men0.30 (0.09)0.57 (0.17)−0.04 (0.06)*1.91 (0.10)0.50 (0.29)*0.58 (0.08)−0.19 (0.14)*0.45 (0.19)−0.17 (0.09)*−2.10 (0.41)1.03 (0.28)−1.07 (0.17)
 Women−1.37 (0.15)1.07 (0.11)−0.88 (0.07)0.35 (0.15)0.65 (0.16)−0.01 (0.06)*−1.35 (0.29)1.18 (0.24)−0.87 (0.12)−4.07 (0.40)1.60 (0.24)−2.25 (0.15)
Age 21 to 59 years            
 Men0.35 (0.12)0.89 (0.08)−0.14 (0.04)1.46 (0.11)0.72 (0.11)0.30 (0.04)−0.02 (0.15)*0.95 (0.10)−0.28 (0.05)−2.41 (0.43)1.32 (0.19)−1.25 (0.14)
 Women−1.64 (0.11)0 (0.06)*−0.62 (0.04)−0.45 (0.07)−0.26 (0.08)−0.09 (0.04)−2.15 (0.14)0.11 (0.12)*−0.81 (0.07)−4.91 (0.32)0.70 (0.16)−2.12 (0.14)
Age 60+ years            
 Men0.50 (0.14)2.79 (0.13)−0.66 (0.06)1.89 (0.29)2.77 (0.20)−0.07 (0.10)*0.01 (0.12)*−2.49 (0.16)−0.75 (0.07)−1.89 (0.41)3.79 (0.34)−2.01 (0.21)
 Women−0.50 (0.14)2.52 (0.11)−0.97 (0.06)0.11 (0.18)*2.28 (0.13)−0.54 (0.07)−0.92 (0.14)2.65 (0.21)−1.23 (0.07)−1.63 (0.57)3.06 (0.32)−1.86 (0.32)

Figure 1 shows a Bland-Altman plot for measured and self-reported BMI. Overall, the difference between measured and self-reported BMI was small, but it slightly increased with increasing average values of these two BMI measures. This shows a greater tendency for self-reported BMI to underestimate true BMI with increasing adiposity.

Figure 1.

Bland-Altman plot of the difference between self-reported and measured BMI vs. the average of these two measures of BMI.

Correlations with Disease Biomarkers

Correlation coefficients for anthropometric measures and disease biomarkers adjusted for sex, race/ethnicity, age group, alcohol consumption, and current cigarette smoking status are presented in Table 4. Overall, similar correlations were seen between the biomarkers and self-reported and measured BMI. For all biological measures, the correlations associated with BIA were either similar to or lower than either BMI correlation value; in the case of SBP, the difference was substantial (Table 4). For the total population, correlations between self-reported BMI and FBG (r = 0.43), HDL (r = −0.53), and SBP (r = 0.54) were not different from those with measured BMI, whereas correlations with TG and CRP were slightly lower for self-reported BMI (Table 4). In contrast, the correlations between BIA measures and disease biomarkers were substantially lower than those for measured BMI. This difference existed in all subgroups except for those <60 years old, for whom the correlation between BIA and CRP was similar to those with measured BMI (r = 0.26). Similar correlations between self-reported and measured BMI with biomarkers were also observed within different measured BMI categories (Table 5).

Table 4.  Adjusted* Pearson correlation for biomarkers and blood measures with self-reported and measured weight, height, and BMI
  • Stratified analyses were adjusted for all confounders excluding the stratification variable.

  • *

    Adjusted for sex, race, alcohol consumption, and current smoking status.

Self-reported BMI0.43−0.530.540.410.260.79
Measured BMI0.43−0.530.540.440.270.82
  Self-reported BMI0.28−0.470.400.390.210.69
  Measured BMI0.29−0.470.410.420.210.72
 Age group (20 to 59 years)      
 Age group (60+ years)      
Table 5.  Adjusted* correlations for biologic measures and anthropometric measures by measured BMI
  • *

    Adjusted for sex, race, and age group.

BMI < 25 kg/m2      
 Self-reported BMI0.34−0.400.370.180.090.73
 Measured BMI0.34−0.400.370.190.090.76
BMI = 25 to 29.9 kg/m2      
 Self-reported BMI0.20−0.540.330.280.150.77
 Measured BMI0.21−0.540.340.330.150.78
BMI ≥ 30 kg/m2      
 Self-reported BMI0.32−0.420.410.300.410.79
 Measured BMI0.33−0.430.410.360.410.82


Epidemiological studies have evolved in the past decades since the Stunkard and Albaum publication (4). With the increasing use of biomarkers of obesity-related disorders in studies, understanding the use of self-reported anthropometric measures is important. This analysis of NHANES III data indicates that self-reported and technician-measured height and weight are highly correlated. We observed underreporting of weight and overreporting of height, causing a slight underestimation of BMI. This was most pronounced in subgroups such as Mexican Americans and older participants. Correlations of self-reported and measured BMI with obesity-related biomarkers were similar not only to each other, but also to those between BIA measures and biomarkers. Although the correlations for self-reported BMI and biomarkers, such as CRP and TG, were slightly lower than for technician measured BMI, these differences do not seem to be of a magnitude that is clinically significant. Both measures showed a moderate to strong positive association between adiposity and the studied biomarkers, and researchers would have drawn similar conclusions regardless of which measure of BMI was used. The latter result agrees with a previous analysis of the NHANES III data that found BIA-estimated body fat was not superior to measured BMI in predicting obesity-related markers (22).

Previous reports of NHANES datasets have found a strong relation between self-reported and measured weight, height, and BMI (6, 8, 10). Similar to Gillum et al. (10), we found that self-reported BMI was under-reported in women and Mexican Americans as a consequence of overreporting height and underreporting weight (10). Variability in the correlations was seen in different ethnic/racial groups. Also, weight and BMI discrepancies were associated with age group and measured BMI, as was noted by Kuczmarski et al. (7) and Villanueva (6). Although many other studies have examined the differences between self-reported and measured values of weight (3, 4, 5, 6, 7, 8, 9, 11, 23), no other studies have examined the correlation of self-reported anthropometric measures and disease biomarkers in nationally representative populations.

NHANES III is a large cross-sectional study of the U.S. population designed for nationally representative samples from the U.S., and results have been widely published. In our analysis, the population includes men and women of diverse ages and racial/ethnic backgrounds. Self-reported measures of weight and height were taken in close proximity to the measured values but were obtained without the participants knowing they would subsequently be measured. Therefore, the self-reported values are similar to what would be obtained in an epidemiological study with only self-reported measures. Different biological measures known to be predictive of obesity-related disorders were used to examine the overall use of different anthropometric measures. The blood levels of biomarkers such as FBG, TGs, and HDLC, as well as SBP, reflect levels of adiposity (24), which makes these variables useful to compare the performance of different anthropometric measures (25).

This study suggests that self-reported BMI can provide sufficiently accurate information in epidemiological studies in which the primary outcomes are disease biomarkers or obesity-related diseases. PBF calculated through BIA was not superior to BMI as a predictor of biological markers known to be associated with adiposity and risk for obesity-related diseases. Although we found self-reported BMI to be highly correlated with measured BMI, some underestimation of true BMI occurs when BMI is based on self-reports. We therefore conclude that in most epidemiological studies of adiposity-related conditions, using self-reported BMI will produce minimal bias for the measure of association, although it can lead to some underestimation of the population prevalence of obesity.


We thank those who dedicated their time to designing and implementing NHANES III. Dr. Hu's research is supported, in part, by an American Heart Association Established Investigator Award.


  • 1

    Nonstandard abbreviations: NHANES III, National Health and Nutirtion Education Study III; BIA, bioelectrical impedance analysis; PBF, percent body fat; FFM, fat-free mass; TBF, total body fat; FBG, fasting blood levels of glucose; TG, triglyceride; HDLC, high-density lipoprotein-cholesterol; SBP, systolic blood pressure; CRP, C-reactive protein; LEP, leptin; CI, confidence interval.

  • The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.