|Home | About | Journals | Submit | Contact Us | Français|
Adiposity is often approximated by body mass index (BMI) in population studies based on self-reported weight and height (kg/m2). However, self-reports tend to underestimate weight and overestimate height, leading to an underestimation of BMI and the prevalence of overweight and obesity. We examined a subgroup of the Multiethnic Cohort Study participants to determine how well self-reported and measured anthropometry correlate with each other, overall and by race/ethnicity, total and abdominal adiposity level, and amount of adulthood weight gain. A cross-sectional sample of 30 Caucasian and 30 Japanese American female cohort participants, between ages 60–65, was selected in such a way the two groups had a similar BMI distribution across the range (18.5–40 kg/m2). Subjects first reported their weight, height, and waist and hip circumferences at home and within several days underwent objective measurements by trained staff and also a whole-body scan of dual energy X-ray absorptiometry (DXA) at a study clinic. The women under-reported their weight by 0.93 kg, waist circumference by 3.95 cm and hip circumference by 0.10 cm and over-reported their height by 0.85 cm. This led to an under-estimation of BMI by 0.67 kg/m2 and waist/hip ratio by 0.04. The effect of misreporting (self-report minus measurement) on BMI and waist/hip ratio was significantly greater in higher BMI groups (p-heterogeneity = 0.007 for BMI, 0.0005 for waist/hip ratio), among women with central obesity (waist circumference > 88 cm; p-heterogeneity = 0.006, 0.01) and among women who had gained higher amounts of weight since age 21 (p-heterogeneity = 0.03, 0.01) compared to their counterparts. A similar trend of greater self-report bias was found among women with higher levels of DXA-based total and abdominal adiposity. We did not observe any heterogeneity in these findings by ethnicity. Our results confirm that a small degree of under-reporting exists in self-reported BMI and waist/hip ratio values, and it tends to increase in women with a larger current body size or history of greater weight gain. Studies are underway to investigate this question in greater depth in men and women from five race/ethnic groups.
Excess body fat is a leading cause of cancer and overall chronic diseases.1,2 In large epidemiologic studies of obesity, body mass index (BMI; weight [kg]/height [meter]2) based on self-reported weight and height is the most commonly used measure of overall body fatness. However, studies have observed that self-reporting tends to underestimate weight and overestimate height.3 Also, the degree of under-reporting weight may be greater at high compared to low levels of BMI, whereas over-reporting height may be greater at older ages.4 We examined a subgroup of women in the Multiethnic Cohort Study to determine how well self-reported and measured anthropometry information correlates with each other, overall and by race/ethnicity, current total and abdominal adiposity, and adulthood weight gain.
The current analysis is based on a cross-sectional study conducted in a small subgroup of the Multiethnic Cohort (MEC) Study. This ongoing prospective study, conducted in Hawai‘i and Los Angeles, consists of over 215,000 men and women aged 45–75 years at recruitment (1993–1996) and of mainly five race/ethnic groups.5 The design of the cross-sectional study was described in detail previously.6 Postmenopausal female MEC participants who were 60–65 years of age as of September 2009 and of either Caucasian or Japanese ethnicity were eligible. Women were excluded if they reported any of the following: current or recent smoking; use of certain medications (chemotherapy, insulin, or weight-loss drugs); substantial weight change in recent months; reported BMI outside the target range (18.5–35 kg/m2); and any soft or metal implants/objects in the body. Of the 218 women contacted, some were ineligible (21%), not reachable (10%), or unwilling to participate (35%). Among the 74 willing and eligible women (34%), we randomly selected 60 women (30 Caucasian, 30 Japanese American [JA]). In order to ensure that Caucasian and JA women would have a similar BMI distribution, equal numbers of women were recruited from the following BMI categories (computed from the weight and height reported on recruitment phone calls) within each ethnic group: 18.5–21.9, 22–24.9, 25–26.9, 27–29.9, and 30–35 kg/m2.6
Participants underwent anthropometric measurements and a whole-body dual energy X-ray absorptiometry (DXA) scan at the University of Hawai‘i Clinical Research Center (CRC). The Institutional Review Board of UH approved the study protocol, and all participants signed an informed consent.
Participants received a questionnaire and a tape measure in the mail and were asked to write in their current weight and height, as well as self-measured waist and hip circumferences using the tape measure, within several days prior to their visit to the CRC clinic. During the clinic visit, trained technicians measured standing height, weight, and waist and hip circumferences. Standing height and weight were measured using a digital stadiometer (Measurement Concepts) and scale (Seca), respectively, with the participant wearing a gown or comparably light clothing. For both self-reports and objective measurements, waist circumference was taken at the navel, and hip circumference was measured at the widest area between the waist and thighs, including the buttocks. Each measurement was taken twice, and a third time if the two measures differed by > 0.1 kg or by > 0.5 cm, after which the average of the two closest values was used in the analysis.
A whole-body DXA scan (GE Lunar Prodigy, Madison, WI) was performed to measure total and regional body fat mass in the trunk, arms and legs, as well as lean soft tissue mass and bone mineral content. The trunk-to-periphery fat mass ratio (TPFR) was calculated as an indicator of abdominal fat distribution by dividing the fat mass in the trunk by the sum of fat mass in the arms and legs.
All analyses were performed using the SAS Statistical Software, version 9.3 (SAS Institute, Inc., Cary, NC), based on the nominal level of α = 0.05 for statistical significance. Two outliers were excluded from the analysis because the self-reported height (177.8 cm) or hip circumference (116.8 cm) was greater than 3 SDs above the measured value (152 cm and 92.1 cm, respectively), leaving a sample size of 58 for analysis. The differences (self-reported value—objective measurement) in weight, height, BMI, waist and hip circumferences, and waist/hip ratio were normally distributed. Mean differences were examined in all study participants and with stratification by ethnicity and level of BMI, central obesity (defined by waist circumference > 88 cm) and weight gain since age 21 (tertiles). Similarly, we examined the mean differences by DXA-based direct measurement of total adiposity (tertiles of percent total fat over total body mass) and abdominal adiposity (below- vs above-median TPFR). All women had reported their weight at age 21 on the MEC baseline questionnaire (1993–1996). Heterogeneity in mis-reporting, ie, variation in the differences between self-reports and measurements, across the ethnicity, current adiposity and weight-gain strata was assessed using the analysis of variance F statistic in a general linear model. Adjusting for age did not influence the results. Graphic illustration is also presented for BMI differences between reported and measured values vs measured BMI, separately for the body size and weight gain categories, and with linear regression lines overlayed. Pearson correlation coefficients were obtained between self-reported and measured anthropometric values, overall and across the above strata.
We previously reported the descriptive characteristics of the cross-sectional study participants and a significant race/ethnic difference in abdominal, visceral and liver fat distribution.6 By study design, the participants had a similar BMI distribution and range (18.5–35 kg/m2 based on self-reports; 18.5–40 kg/m2 based on measurements) in both ethnic groups (Table 1). Overall, the women under-reported their weight by 0.93 kg (2.05 lbs) and over-reported their height by 0.85 cm (0.33 inches), which led to an under-estimation of BMI by 0.67 kg/m2 (range −5.97, 3.27). They also under-reported waist (by 3.95 cm or 1.56 inches) and hip (by 0.10 cm or 0.04 inches) circumferences, yielding a slightly under-estimated waist/hip ratio (by 0.04; range −0.16, 0.10). There was no significant variation in misreport between the two ethnic groups for any of the anthropometric indicators. The under-estimations of self-reported weight and waist circumference were significantly greater for the higher BMI group, among women with central obesity (waist circumference >88 cm), and among women who gained higher amounts of weight since age 21. These biased self-reports resulted in increasing under-estimations of BMI and waist/hip ratio among heavier women (p-heterogeneity = 0.007 and 0.0005, respectively, for BMI and waist/hip ratio), women with central obesity (p-heterogeneity = 0.006, 0.01), and women who experienced more adulthood weight gain (p-heterogeneity = 0.03, 0.01) (Table 1). This is also illustrated in Figure 1 for BMI. Greater self-report bias was similarly observed at higher levels of DXA-based total adiposity (percent total fat; p-heterogeneity = 0.008 and 0.006 for BMI and waist/hip ratio, respectively) and abdominal adiposity (TPFR; p-heterogeneity = 0.09 and 0.03) (Table 1).
The correlation between self-reported and measured values was high for weight and height (r > 0.75) and for resultant BMI (r ≥ 0.90), except for BMI among overweight (r = 0.58) and obese women (r = 0.73) (Table 2). Correlations were more moderate for waist and hip circumferences (r > 0.70) and waist/hip ratio (r range 0.62–0.81).
Our findings confirm that self-reports lead to an under-estimation of body weight and BMI in women. Also, we observed a similar trend in waist circumference reports and subsequent waist/hip ratio estimates. The overall magnitude of under-estimation in this study population was relatively small at 0.67 kg/m2 for BMI and 0.04 for waist/hip ratio. Also, the correlation between self-reported and measured values was high, indicating that investigations of reported anthropometry in relation to cancer or other health outcomes may be minimally affected in ranked, categorical analyses. However, the magnitude of self-report bias was greater among overweight and obese women, women with greater abdominal adiposity, and women who had gained more than 5 lbs compared to their recalled weight at age 21. Such differential extent of bias would attenuate the risk estimates of the true association between excess adiposity and disease outcomes, and highlight the need to account for, in analysis, the factors that may modify the degree of self-report bias.
The comprehensive anthropometry data collected in this study allowed us to examine self-report bias in waist size, as well as BMI estimation, and to compare the misreporting by the level of DXA-based direct measurements of total and abdominal adiposity. Our study participants, as in previous studies that compared self-reported and measured anthropometry, might have reported their weight, height, and circumferences more accurately than they would have if they had not agreed to a follow-up clinic visit. We did not observe any significant difference in misreport within the narrow study age range (60–65 years) or between the two ethnic groups that had balanced BMI distributions by study design. Therefore, our study may not be suitable to test the heterogeneity in self-report bias by age or ethnicity.
We are currently expanding the cross-sectional investigation to 2,000 men and women of all five ethnic groups in the MEC, including Native Hawaiians, African Americans, and Latinos, and will be able to examine in greater depth whether bias in self-report varies among the sex-ethnic groups, as well as with increasing levels of adiposity and adulthood weight gain. This study may also enable us to identify some other factors that are associated with reporting error, including eating behaviors, fitness and etc.
Funding for this study was provided in part by the National Cancer Institute for the Multiethnic Cohort (R37 CA54281: LNK), for University of Hawai‘i Cancer Center seed funding (P30 CA071789: Carl Vogel) and for the “Obesity, Body Fat Distribution and Cancer Risk in the Multiethnic Cohort” study (P01 CA168530: LLM), and by the National Center for Research Resources at NIH for the University of Hawai‘i Clinical Research Center (P20 RR11091: Jerris Hedges). We thank the study participants and the dedicated staff at the University of Hawai‘i Cancer Center (Karin Koga, Eugene Okiyama, Naomi Hee, Janice Nako-Piburn, Janine Abe, Wileen Mau, Maj Earle) and Clinical Research Center (Sara Murakami, Jane Yakuma, Patty Iwamoto).
The authors declared no conflict of interest.