Thiet

Nutritional Methodology

Self-Reported and Technician-Measured Waist Circumferences Differ in Middle-Aged Men and Women Janne Bigaard,*†1 Iben Spanggaard,* Birthe Lykke Thomsen,* Kim Overvad,** and Anne Tjønneland* *Institute of Cancer Epidemiology, The Danish Cancer Society, Copenhagen, Denmark; †Danish Epidemiology Science Centre and Research Unit for Dietary Studies at the Institute of Preventive Medicine, Copenhagen University Hospital, Copenhagen, Denmark; and **Department of Clinical Epidemiology, Aalborg Hospital, and Aarhus University Hospital, Aalborg, Denmark

KEY WORDS:

●

waist circumference

●

self-reported

●

The worldwide obesity epidemic (1) has led to a demand for simple obesity indicators to monitor the development and prevalence of obesity. The WHO recommended the use of the BMI (weight divided by height squared) as an indicator of obesity (1). BMI does not directly reflect the distribution of fat in the body (2). In the mid 1990s, waist circumference was introduced as an independent risk indicator identifying individuals with need for weight management (3) and an accumulation of cardiovascular risk factors (4,5). Waist circumference was implemented in clinical guidelines for the treatment of obesity (1,6,7). The evidence for waist circumference as a risk indicator is accumulating (8 –12). Earlier validation studies of self-reported vs. technicianmeasured waist circumference were published in the early 1990s, before identification of the obesity epidemic; most studies used correlation and comparison of the mean differences (13–21), although this approach may be misleading. A high correlation between 2 methods occurs if 1 method always measures twice as much as the other because correlation assesses only the closeness to a linear relation between the methods and not agreement or whether the methods can be

umbilicus

●

middle age

used interchangeably (22,23). The degree of correlation does indicate how well 1 method works as a proxy for the other as a linear covariate or as an outcome in a regression analysis aimed at establishing statistical significance. In this case, a constant displacement does not matter, but a slope different from 1 in the linear relation between the 2 measurement methods would bias the estimated association in future regression analyses accordingly. In categorical analyses using prespecified limits for the categories, a constant displacement would lead to some misclassification. Follow-up data of waist circumference and weight were collected for the Danish Diet, Cancer and Health cohort to evaluate the importance of changes in body size and fat distribution for the development of cancer and other diseases. At the 5-y follow up, we collected self-reported values of weight and waist circumference at the level of the umbilicus. At baseline, laboratory technicians measured weight, height, circumference at the natural waist, and hip circumference. We changed the body site for measurement of waist circumference to the level of the umbilicus at follow-up to simplify the measurement instructions for participants using umbilicus as a body mark for measurement. The aim of this study was to investigate agreement 1) between self-reported and technician-measured waist circum-

1 To whom correspondence should be addressed. E-mail: [email protected].

0022-3166/05 $8.00 © 2005 American Society for Nutritional Sciences. Manuscript received 2 February 2005. Initial review completed 8 March 2005. Revision accepted 1 July 2005. 2263

Downloaded from jn.nutrition.org by on April 18, 2009

ABSTRACT We investigated the agreement between 1) self-reported and technician-measured waist circumference at the level of the umbilicus, 2) circumference measured at the level of the umbilicus and halfway between the lower rib and the iliac crest (the natural waist), and 3) self-reported circumference at the level of the umbilicus and technician-measured circumference at the natural waist. At follow-up in the Danish “Diet, Cancer and Health” study, we recruited 176 men and 240 women for a validation study. Bland-Altman plots were used to evaluate agreement among measurement sites. Multiple regression was used to identify variables explaining the difference between measurements. The participants underestimated their waist circumference; the mean differences were ⫺1.6 cm (95% CI: ⫺2.4 cm, ⫺0.8 cm) in men and ⫺3.0 cm (95% CI: ⫺3.8 cm, ⫺2.3 cm) in women. Limits of agreement were from ⫺11.9 to ⫹8.7 cm among men and ⫺14.9 to ⫹8.9 cm among women. High BMI and large baseline waist circumference were associated with a larger degree of underreporting. Waist circumference measured at the level of the umbilicus was larger than at the natural waist; the mean differences were ⫹0.7 cm (95% CI: ⫹0.4 cm, ⫹1.1 cm) in men and ⫹5.0 cm (95% CI: ⫹4.4 cm, ⫹5.6 cm) in women. The self-reported waist circumference at the level of the umbilicus was correlated with the technician-measured circumference at the natural waist. The circumference at the natural waist was overestimated for women, depending on baseline waist circumference, and slightly underestimated for men, depending on baseline BMI. J. Nutr. 135: 2263–2270, 2005.

BIGAARD ET AL.

2264

ference at the level of the umbilicus, 2) between waist circumference measured at the level of the umbilicus and at the narrowest part between the lower rib and the iliac crest (the natural waist) and 3) between self-reported waist circumference at the level of the umbilicus and technician-measured waist circumference at the natural waist. Further, we investigated the influence of the body size and the body shape on the possible differences. We considered technician-measured BMI and baseline body shape for differences 1) and 2). Difference 3) is the important difference in future studies evaluating the association between changes in waist circumference and different exposures in the total cohort. We therefore investigated body size measurements available for the total cohort, namely, baseline technician-measured BMI and baseline body shape measurements. SUBJECTS AND METHODS

RESULTS

FIGURE 1 Flow chart of the recruitment of participants for the validation study in 2 rounds. WC, waist circumference.

In round 1 of the recruitment (May–June 2001), 582 people were invited after random selection stratified according to sex and age group (55–59, 60 – 64, and ⱖ65 y) among 9647 potential participants who had returned their follow-up questionnaire (Fig. 1). Of the 582 invited, 256 (44%) visited the validation clinic. In round 2 of the recruitment (November– December 2001), 258 people were invited after random selection stratified according to sex, age group, and BMI (⬍18.5, 18.5–24.9, 25–29.9, ⱖ30 kg/m2) among the 1927 potential participants who had returned their follow-up questionnaire within the previous 2 mo (Fig. 1). However, to recruit a group of women with BMI ⬍ 18.5 kg/m2, we allowed a time interval of up to 6 mo after return of their follow-up questionnaire. It was not possible to recruit a similar group among men, because very few had a BMI ⬍ 18.5 kg/m2. Of the 258 invited, 160 (62%) visited the validation clinic. A total of 416 (176 men and 240 women) were recruited for the validation study; however, 8 were excluded from analyses using self-reported

Downloaded from jn.nutrition.org by on April 18, 2009

Between December 1993 and May 1997, 27,178 men (33.6% of total number eligible) and 29,875 women (37.5% of total number eligible) aged 50 – 64 y were recruited for the Danish prospective study “Diet, Cancer and Health.” All participants in the cohort were born in Denmark and were inhabitants of the greater Copenhagen or Aarhus areas. Follow-up. All participants received a postal questionnaire between September 1999 and August 2002, ⬃5 y after baseline data collection. In the questionnaire, participants were instructed to report their weight in kilograms and their waist circumference in centimeters measured at the level of the umbilicus using an enclosed paper measuring tape. Validation study. In connection with the data collection for the 5-y follow-up, participants were recruited for a validation study. Potential participants were recruited in 2 rounds (Fig. 1). In round 1 of the recruitment (June–August 2001), potential participants returned their follow-up questionnaire between October 1, 1999 and June 1, 2001 and 44% participated (256 of the 582 invited). All participants were inhabitants of Copenhagen County. In round 2 of the recruitment (November–December 2001), potential participants returned their follow-up questionnaire between August 1 and Octo-

ber 1, 2001 and 62% participated (160 of the 258 invited). All participants were inhabitants of Copenhagen County. The study was conducted in accordance with the Helsinki Declaration II and approved by the Ethical Committees on Human Studies in Copenhagen and Aarhus municipalities. Measurements. Measurements were performed in a study clinic situated in Copenhagen. Measurements were performed in accordance with instructions for the baseline data collection. Height was measured with the participants standing without shoes and was recorded to the nearest 0.5 cm. Weight was measured using a digital scale, with the participants wearing light clothing or underwear, and was recorded to the nearest 100 g. Waist circumference was measured at the narrowest part between the lower rib and the iliac crest (the natural waist) or, in the case of an indeterminable waist narrowing, halfway between the lower rib and the iliac crest, and was recorded to the nearest 0.5 cm. Additionally, waist circumference was measured at the level of the umbilicus and recorded to the nearest 0.5 cm. Two investigators, the first and the second authors of this paper, performed the anthropometric measurements, except for 7 participants measured by a third technician. The interobserver variation of the waist circumference measured at the level of umbilicus by the 2 investigators was investigated in 47 participants (22 men and 25 women). Statistical methods. The different measurements of waist circumferences were compared by paired t tests. For illustrative purposes, we calculated Spearman’s correlation between measurements. Agreements between measurements were illustrated in Bland-Altman plots (22,23) in which the difference between the 2 measurements are plotted against the mean of the 2 measurements. Limits of agreements were calculated as the mean difference ⫾ 2 SD, and the 95% CI was calculated using Bland and Altman’s method to assess the precision of these limits (23). Multiple regression was used to identify variables explaining difference 1: self-reported minus technician-measured waist circumference at the level of the umbilicus; difference 2: waist circumference measured at the level of the umbilicus minus circumference measured at the natural waist; and difference 3: self-reported waist circumference at the level of umbilicus minus technician-measured circumference at the natural waist. We investigated the following variables as covariates for the agreement between measurements; difference 1: months between measurements, body size as current BMI (current weight divided by current height squared, weight and height measured by the technicians in the validation clinic), and body shape (waist and hip circumference measured at baseline); difference 2: the same variables except “months between measurements,” which was omitted because measurements took place the same day; and difference 3: months between measurements, current BMI (weight and height measured by the technicians in the validation clinic), baseline BMI (baseline weight divided by baseline height squared), and body shape (waist and hip circumference measured at baseline). The analyses were performed using SAS (SAS Institute, version 8).

DIFFERENCE BETWEEN MEASUREMENTS OF WAIST CIRCUMFERENCE

2265

TABLE 1 Age and body measurements among participants in the Danish Diet, Cancer and Health validation study at baseline and follow-up Men (n ⫽ 176) Mean

Min

Median

Max

Mean

Min

Median

Max

56.9 83.7 177.3 26.6 95.8 100.1

50.2 54.3 159.0 19.5 73.0 83.0

56.5 82.6 177.0 26.3 94.0 100.0

64.8 117.6 199.5 38.7 130.0 126.0

56.5 66.5 164.2 24.7 80.0 99.8

50.2 39.0 149.0 16.8 60.0 79.0

56.2 65.6 164.0 24.5 79.0 99.0

65.2 114.0 179.0 40.9 119.0 130.0

63.0

56.1

62.6

71.3

62.4

55.7

61.9

72.1

83.9 26.7 99.3

52.0 17.4 78.0

83.0 26.2 98.0

126.0 40.7 134.0

66.8 24.8 88.1

40.0 16.6 60.0

65.0 24.3 87.0

117.0 42.0 131.0

84.7 176.8 27.1 101.1 100.4

55.5 159.0 18.9 75.5 76.0

84.0 177.0 26.6 100.0 99.0

126.7 200.0 40.4 133.0 133.0

68.1 163.7 25.5 91.1 86.1

40.2 148.0 16.3 64.0 60.0

67.4 164.0 25.0 91.0 85.0

114.3 180.0 41.5 138.0 125.0

waist circumference because they did not report waist circumference in the follow-up questionnaire, leaving 408 (173 men and 235 women) eligible for those analyses. Descriptions of age and body measurements at baseline and follow-up, the variables used in the analyses, are reported in Table 1. The number of months between the return of the questionnaire (self-reported measurement of waist circumference) and the visit to the validation clinic (technician-measured waist circumference) was ⬍6 mo for 53% of the men and 75% of the women (Table 2). In round 2 of invitation, all men and 89% of the women had time intervals ⬍ 3 mo. The self-reported and technician-measured waist circumferences were highly correlated. Spearman’s correlation coefficient was 0.87 in men and 0.88 in women. The BlandAltman plots show the mean differences and limits of agreement between self-reported and technician-measured waist circumference at the level of the umbilicus (Fig. 2). Negative differences between measurements indicate that participants underestimated their waist circumference and positive differences that they overestimated the waist circumference. The mean difference (solid line) was ⫺1.6 cm (95% CI: TABLE 2 Participants in the validation study distributed according to months between the return of the questionnaire and the visit to the validation clinic Months, n

Men (n ⫽ 173)

Women (n ⫽ 235) % (n)

ⱕ3 3–6 6–9 9–12 12–15 ⬎15

31.2 (54) 22.0 (38) 8.7 (15) 30.1 (52) 6.4 (11) 1.7 (3)

57.0 (134) 18.3 (43) 4.7 (11) 15.7 (37) 3.4 (8) 0.9 (2)

⫺2.4 cm, ⫺0.8 cm) among men (Fig. 2, upper panel) and the 95% limits of agreement (dotted lines) were ⫺11.9 to ⫹8.7 cm (95% CI: lower limit ⫺12.3 cm, ⫺11.4 cm; upper limit ⫹8.3 cm, ⫹9.1 cm). In women (Fig. 2, lower panel) the mean difference (solid line) was ⫺3.0 cm (95% CI: ⫺3.8 cm, ⫺2.3 cm) and limits of agreement (dotted lines) were ⫺14.9 to ⫹8.9 cm (95% CI: lower limit ⫺15.6 cm, ⫺14.3 cm; upper limit ⫹8.2 cm, ⫹9.5 cm). The negative correlation between means and differences was not significant for either men (P ⫽ 0.44) or women (P ⫽ 0.11). For 40% of the men and 27% of the women, the waist circumference reported was within 2 cm difference in either direction of the technician-measured waist circumference. A higher proportion, 42% of the men and 57% of the women, underestimated their waist circumference by ⬎2 cm, whereas overestimation was less common (18% men and 16% women). Spearman’s correlation coefficients between the natural waist and the waist circumference measured at the level of the umbilicus were very high, 0.98 in men and 0.93 in women. Negative differences between measurements in the BlandAltman plots indicate that waist circumference measured at the level of the umbilicus was smaller than measured at the natural waist and positive differences that waist circumference measured at the level of the umbilicus was larger (Fig. 3). The mean difference (solid line) was ⫹0.7 cm (95% CI: ⫹0.4 cm, ⫹1.1 cm) among men (Fig. 3, upper panel) and the 95% limits of agreement (dotted lines) were ⫺3.8 to ⫹5.3 cm (95% CI: lower limit ⫺4.0 cm, ⫺3.6 cm; upper limit ⫹5.1 cm, ⫹5.5 cm). In women (Fig. 3, lower panel), the mean difference (solid line) was ⫹5.0 cm (95% CI: ⫹4.4 cm, ⫹5.6 cm) and limits of agreement (dotted lines) were ⫺4.5 to ⫹14.5 cm (95% CI: lower limit ⫺5.6 cm, ⫺3.4 cm; upper limit ⫹13.4 cm, ⫹15.6 cm). The correlation between means and differences was negative for men (P ⫽ 0.01) and positive for women (P ⫽ 0.002). For 72% of the men and 30% of the women, the difference between the measurements of the waist circumference was within 2 cm in either direction. The waist circumference at the level of the umbilicus was at least 2 cm larger than the

Downloaded from jn.nutrition.org by on April 18, 2009

Baseline Age, y Weight, kg Height, cm BMI, kg/m2 Waist circumference (WC), cm Hip circumference, cm Follow-up Age, y Self-reported Weight, kg BMI, kg/m2 WC, cm Technician-measured Weight, kg Height, cm BMI, kg/m2 WC at the umbilicus, cm WC at the natural waist, cm

Women (n ⫽ 240)

2266

BIGAARD ET AL.

(solid line) was ⫺0.8 cm (95% CI: ⫺1.6 cm, ⫹0.007 cm) among men (Fig. 4, upper panel) and the 95% limits of agreement (dotted lines) were ⫺11.3 to ⫹9.7 cm (95% CI: lower limit ⫺11.1 cm, ⫺11.5 cm; upper limit ⫹9.5 cm, ⫹9.9 cm). In women (Fig. 4, lower panel), the mean difference (solid line) was ⫹2.1 cm (95% CI: ⫹1.3 cm, ⫹2.9 cm) and limits of agreement (dotted lines) were ⫺10.5 to ⫹14.6 cm (95% CI: lower limit ⫺10.0 cm, ⫺11.0 cm; upper limit ⫹14.2 cm, ⫹15.1 cm). The means and the differences were negatively correlated in men (P ⫽ 0.03) but not in women (P ⫽ 0.71). The difference between self-reported and technician-measured waist circumference was related to current BMI and baseline body shape adjusted for months between measure-

Downloaded from jn.nutrition.org by on April 18, 2009

FIGURE 2 Bland-Altman plot of the difference in men (upper panel) and women (lower panel) between self-reported and technicianmeasured waist circumference (WC) in cm at the level of the umbilicus plotted against the mean of self-reported and technician-measured waist circumferences. Horizontal lines represent the mean difference and 95% limits of agreement. The thin line shows the regression of the differences on the means.

natural waist in 21% of the men and 67% of the women, and only a few participants had a smaller waist circumference at the level of the umbilicus than at the natural waist (7% men and 3% women). Spearman correlation coefficients between self-reported waist measured at the level of the umbilicus and the technician-measured waist circumference at the natural waist were high, 0.88 in men and 0.86 in women. Negative differences between measurements in the Bland-Altman plots indicate that self-reported waist circumference was smaller than technician-measured circumference at the natural waist; positive differences indicate that the technician-measured circumference at the natural waist was smallest. The mean difference

FIGURE 3 Bland-Altman plot of the difference in men (upper panel) and women (lower panel) between technician-measured waist circumferences (WC) in cm at the level of the umbilicus and halfway between the lower rib and the iliac crest plotted against the mean of the 2 waist circumferences. Horizontal lines represent the mean difference and 95% limits of agreement. The thin line shows the regression of the differences on the means.

DIFFERENCE BETWEEN MEASUREMENTS OF WAIST CIRCUMFERENCE

ments (Table 3). The first column presents the estimates for separate regressions on body size (current BMI) and body shape (mutually adjusted waist and hip circumference). The participants with higher current BMI reported a smaller waist circumference relative to the technician-measured waist circumference among both men (⫺0.36 cm; 95% CI: ⫺0.54, ⫺0.18 per 1 kg 䡠 m2) and women (⫺0.26 cm; 95% CI: ⫺0.42, ⫺0.10 per 1 kg 䡠 m2). Either of the 2 body shape measurements could be omitted (all P ⱖ 0.18), but not both for women (P ⫽ 0.001), showing that the 2 circumferences captured the same underlying aspect equally well for women. When current BMI and baseline body shape measurements, waist and hip circumference, were mutually adjusted, current BMI was most

important in men but not in women. The included variables explained only 10% of the variation in men and 7% in women. With higher current BMI, men and women underestimated their waist circumference more (Table 4). Only normal weight men and underweight women (BMI ⬍ 18.5 kg/m2) did not underreport their waist circumference. Men with baseline waist circumference above the 3rd quartile (⬎102 cm) and women with baseline waist circumference above the 1st quartile (⬎80 cm) significantly underreported their waist circumference at follow-up (Table 4). The difference between the waist circumference measured at the level of the umbilicus and the natural waist was related to current body size and baseline body shape (Table 5). Column 1 shows the separate estimates for body size (current BMI) and body shape (mutually adjusted waist and hip circumferences). The waist circumference measured at the level of the umbilicus was relatively smaller than the waist circumference at the natural waist for men with higher BMI (⫺0.11 cm; 95% CI: ⫺0.19, ⫺0.03 per 1 kg 䡠 m2), whereas the women showed an insignificant association in the opposite direction (0.11 cm; 95% CI: ⫺0.02, 0.24 per 1 kg 䡠 m2)]. For body shape, the associations with the waist and hip circumferences at baseline were in opposite directions but of similar magnitude (test of equality; men P ⫽ 0.59, women P ⫽ 0.04), indicating that the important aspect was the difference between the 2 circumferences. The larger the difference between waist circumference and hip circumference at baseline, the larger was the difference between waist circumference at the 2 body sites at follow-up. For each additional centimeter the hip was larger than the waist circumference at baseline, the waist circumference at the level of the umbilicus was ⬃0.12 cm (95% CI: 0.06, 0.18) larger for men and 0.11 cm (95% CI: 0.02, 0.24) larger for women. When mutually adjusted (column 2), body size (current BMI) was not significant, whereas body shape remained significant for the difference in both men and women. The variables included explained only 10% of the variation in men and 5% in women. The difference between self-reported waist circumference at the level of the umbilicus and the technician-measured natural waist was related to the body size and body shape measurements available for the total cohort (Table 6). Column 1 shows the separate estimates for body size (self-reported BMI and baseline BMI) and body shape (mutually adjusted waist and hip circumferences). Column 2 indicates that BMI calculated from baseline weight and height measured by technicians was more accurate than a self-reported BMI calculated from self-reported weight at follow-up and baseline height for both men and women. Column 3 includes body size and body shape, mutually adjusted. Body size (BMI at baseline) remained significant for underreporting in men, whereas baseline waist circumference was significant in women. The variables included explained only 8% of the variation in men and 5% in women We found no systematic interobserver differences in the measurement of the waist circumference at the level of the umbilicus. In men, the mean difference between observers was ⫺0.2 cm (95% CI: ⫺0.8, 0.4) and the limits of agreement were ⫺3.0 cm to ⫹2.6 cm (95% CI: lower limit ⫺4.0 cm, ⫹1.6 cm; upper limit ⫺2.0 cm, ⫹3.6 cm). In women, the mean difference between observers was ⫺0.2 cm (95% CI: ⫺1.2, ⫹0.9) and limits of agreement were ⫺5.6 cm to ⫹5.3 cm (95% CI: lower limit ⫺7.3 cm, ⫹3.6 cm; upper limit ⫺3.9 cm, ⫹7.0 cm).

Downloaded from jn.nutrition.org by on April 18, 2009

FIGURE 4 Bland-Altman plot of the difference in men (upper panel) and women (lower panel) between self-reported waist circumference (WC) in cm at the level of the umbilicus and technician-measured waist circumference halfway between the lower rib and the iliac crest plotted against the mean of the 2 measurements. Horizontal lines represent the mean difference and 95% limits of agreement. The thin line shows the regression of the differences on the means.

2267

BIGAARD ET AL.

2268

TABLE 3 Regression analyses of self-reported waist circumference (WC) at the level of the umbilicus minus technician-measured WC at the level of the umbilicus on current body size and baseline body shape Separate models1

Men (n ⫽ 173) Current BMI, kg/m2 Body shape WC,2 cm HC,2 cm Women (n ⫽ 235) Current BMI, kg/m2 Body shape WC,2 cm HC,2 cm

Estimate

95% CI

⫺0.36

⫺0.54, ⫺0.18

0.004 ⫺0.13

Mutually adjusted model1 P-value

Estimate

95% CI

P-value

0.0001

⫺0.56

⫺0.87, ⫺0.26

0.0004

⫺0.14, 0.15 ⫺0.34, 0.08

0.96 0.083 0.21

0.13 ⫺0.04

⫺0.03, 0.28 ⫺0.25, 0.17

0.10 0.74

⫺0.26

⫺0.42, ⫺0.10

0.002

⫺0.01

⫺0.32, 0.29

0.93

⫺0.08 ⫺0.08

⫺0.20, 0.04 ⫺0.22, 0.07

0.18 0.0013 0.30

⫺0.08 ⫺0.07

⫺0.21, 0.06 ⫺0.23, 0.09

0.27 0.37

0.413

0.183

Time (mo) between measurements was included in the models; men: P ⫽ 0.72, women: P ⫽ 0.12. WC and hip circumference (HC) were mutually adjusted. 3 Simultaneous test of WC and HC. 1 2

the agreement between the measurements, but the correlation depends on the range and the distribution of the variables; furthermore, correlation ignores any systematic bias between the 2 measurements (23). Bland and Altman defined limits of agreement as the interval including 95% of the observed differences (22,23). It is a clinical, not a statistical decision, how narrow these limits of agreement should be to consider the agreement between the 2 methods to be good (23); our findings are examples of high correlation between measurements but poor agreement because of systematic errors. Earlier validation studies of the waist circumference based the evaluation of the agreement between the measurements on the correlation, whereas the limits of agreement were not calculated (13–21). The few studies using limits of agreement to evaluate agreement between measurements also found wide limits of agreement (poor agreement). In the Glasgow MONICA study (20), the mean difference between self-reported and technician-measured waist circumference was ⫺6.7 cm in men and the limits of agreement were from ⫺21.0 cm to ⫹7.7 cm. In women, the mean difference was ⫺4.3 cm with limits of agreement from ⫺21.1 to ⫹12.3 cm. In the same

These middle-aged men and women underestimated their waist circumference. Women underestimated their waist circumference more than men, and the underestimation depended on body size so that the underestimation was stronger with higher baseline BMI and waist circumference in both sexes. Waist circumference measured at the level of the umbilicus was larger than that measured midway between the lower rib and the iliac crest (the natural waist). The difference between the waist circumference measurements at the 2 body sites was largest in women and depended on body shape so that the difference increased with a larger difference between hip and waist circumference at baseline. The self-reported waist circumference at the level of the umbilicus was correlated with the technician-measured circumference at the natural waist. The circumference at the natural waist was overestimated for women depending on their baseline waist circumference and slightly underestimated for men depending on baseline BMI. Good agreement was found between measurements by the 2 investigators in the validation clinic. Studies frequently use correlation coefficients to evaluate

TABLE 4 The mean difference (Dif) between self-reported and technician-measured waist circumference (WC) at the level of the umbilicus according to WHO’s categories of current BMI and baseline WC quartiles Men n ⫽ 173

Dif1 (95% CI)

n ⫽ 235

Dif1 (95% CI)

Underweight: BMI ⬍ 18.5 Normal weight: BMI 18.5–24.99 Overweight: BMI 25–29.99 Obese: BMI ⱖ30

0 56 74 43

0.2 (⫺1.1, 1.4) ⫺1.8 (⫺2.9, ⫺0.7) ⫺3.5 (⫺5.3, ⫺1.6)

18 99 74 44

⫺0.0 (⫺2.6, 2.5) ⫺2.0 (⫺3.2, ⫺0.8) ⫺3.7 (⫺5.1, ⫺2.3) ⫺5.5 (⫺7.1, ⫺3.9)

WC groups2 Lowest Lower middle Upper middle Highest

56 38 34 45

⫺1.0 (⫺2.4, 0.3) ⫺1.4 (⫺3.0, 0.3) ⫺1.4 (⫺3.2, 0.4) ⫺2.6 (⫺4.3, ⫺0.9)

73 56 53 53

⫺1.2 (⫺2.4, 0.1) ⫺2.3 (⫺4.0, ⫺0.5) ⫺5.0 (⫺6.5, ⫺3.4) ⫺4.5 (⫺6.1, ⫺2.9)

Current BMI, kg/m2

1 2

Women

Self-reported WC minus technician-measured WC, both at the level of the umbilicus. Baseline WC quartile groups. Quartiles among total cohort, 57,053 participants: 89, 95, 102 cm in men and 74, 80, 88 cm in women.

Downloaded from jn.nutrition.org by on April 18, 2009

DISCUSSION

DIFFERENCE BETWEEN MEASUREMENTS OF WAIST CIRCUMFERENCE

2269

TABLE 5 Regression analyses of technician-measured waist circumference (WC) at the level of the umbilicus minus technician-measured WC halfway between the lower rib and the iliac crest on current BMI and baseline body shape Separate models1

Men (n ⫽ 176) Current BMI, kg/m2 Baseline body shape WC,2 cm HC,2 cm Women (n ⫽ 240) Current BMI, kg/m2 Baseline body shape WC,2 cm HC,2 cm

Estimate

95% CI

⫺0.11

⫺0.19, ⫺0.03

⫺0.12 0.14

⫺0.18, ⫺0.06 0.05, 0.23

0.11 ⫺0.11 0.18

Mutually adjusted model1 P-value

Estimate

0.01 ⬍0.0001 0.00023 0.0025

⫺0.02, 0.24

0.10

⫺0.20, ⫺0.02 0.07, 0.30

0.02 0.013 0.002

95% CI

P-value

⫺0.10

⫺0.24, 0.03

0.13

⫺0.10 0.16

⫺0.17, ⫺0.03 0.06, 0.25

0.003 0.0023 0.0009

⫺0.06, 0.42

0.14

⫺0.26, ⫺0.04 0.02, 0.27

0.007 0.03

0.18 ⫺0.15 0.15

0.013

1

BMI based on technician-measured weight and height at the validation clinic. WC and hip circumference (HC) were mutually adjusted. 3 Simultaneous test of WC and HC. 2

Some studies included pictorial measuring instructions (16,18), or the participants were asked to seek help from another person (13). Our questionnaire included short, simple written instructions: “Measure the waistline at the level of the navel, exhale, and mark the circumference on the measuring tape. Note the circumference in rounded centimeters.” We had expected that these middle-aged individuals would not be so concerned about their body size and thus would report their waist circumference in an unbiased manner. This assumption was incorrect. An English validation study of weight in women aged 60 –79 y had a similar hypothesis, but elderly women in that study also underestimated their weight (25). The finding that waist circumference is larger at the level of the umbilicus agrees with another study showing that the waist circumference measured at the narrowest waist was smaller than the waist circumference measured immediately below the

TABLE 6 Regression analyses of self-reported waist circumference (WC) at the level of the umbilicus minus technician-measured narrow WC halfway between the lower rib and the iliac crest on self-reported body size, baseline body size, and body shape Separate models1

Men (n ⫽ 173) BMI,2 kg/m2 Self-reported Baseline Body shape WC,3 cm HC,3 cm Women (n ⫽ 235) BMI,2 kg/m2 Self-reported Baseline Body shape WC,3 cm HC,3 cm

Mutually adjusted model1

Estimate

95% CI

P-value

⫺0.36 ⫺0.37

⫺0.55, ⫺0.16 ⫺0.58, ⫺0.17

0.0005 0.0003

⫺0.10 ⫺0.02

⫺0.25, 0.04 ⫺0.23, 0.20

0.17 0.024 0.88

⫺0.17 ⫺0.26

⫺0.36, 0.01 ⫺0.45, ⫺0.07

0.07 0.008

⫺0.19 ⫺0.11

⫺0.32, ⫺0.07 ⫺0.05, 0.26

0.003 0.0034 0.18

Estimate

⫺0.14 ⫺0.25

0.41 ⫺0.66

95% CI

⫺0.59, 0.32 ⫺0.71, 0.22

⫺0.07, 0.89 ⫺1.15, ⫺0.16

P-value

Estimate

95% CI

0.56 0.29

⫺0.58

⫺1.04, ⫺0.13

0.01

0.04 0.08

⫺0.14, 0.22 ⫺0.15, 0.30

0.65 0.50

⫺0.12

⫺0.61, 0.38

0.65

⫺0.17 0.13

⫺0.33, 0.00 ⫺0.06, 0.32

0.05 0.17

0.09 0.01

Time (mo) between measurements was included in the models; men: P ⫽ 0.84, women: P ⫽ 0.87. BMI based on self-reported weight in the follow-up questionnaire and height measured at baseline. 3 WC and hip circumference (HC) were mutually adjusted. 4 Simultaneous test of WC and HC. 1 2

Mutually adjusted model1 P-value

0.574

0.084

Downloaded from jn.nutrition.org by on April 18, 2009

study, the use of a specially designed “Waist Watcher” measuring tape gave a closer agreement; in men the mean difference was ⫺0.5 cm with limits of agreements from ⫺6.2 cm to ⫹5.2 cm, and in women the mean difference was ⫺0.4 cm with limits of agreement from ⫺5.0 cm to ⫹4.2 cm (20). In the Health Professionals’ Follow-up Study, close agreement was found; the mean difference was ⫹0.14 cm and the limits of agreement were from ⫺6.02 cm to ⫹6.30 cm (24). From the same cohort, an earlier publication found a mean difference of ⫹0.36 cm in men and we calculated the corresponding limits of agreement to be from ⫺5.1 cm to ⫹6.9 cm; in women, the mean difference was ⫺0.05 cm with limits of agreement from ⫺10.2 cm to ⫹10.0 cm (participants in the Nurses’ Health Study) (15). These findings are in accordance with our results considering that participants in both the Males Professionals’ Follow-up Study (15,24) and the Nurses’ Health Study (15) were not lay people and 1 study used a specially designed measuring tape (20).

BIGAARD ET AL.

2270

ACKNOWLEDGMENTS We thank programmer Katja Boll and secretary Jytte Fogh Larsen for their work with the database and data collection.

LITERATURE CITED 1. World Health Organization (2000) Obesity: Preventing and Managing the Global Epidemic. Report of a WHO consultation. WHO, Geneva, Switzerland. 2. Baumgartner, R. N., Heymsfield, S. B. & Roche, A. F. (1995) Human body composition and the epidemiology of chronic disease. Obes. Res. 3: 73–95. 3. Lean, M. E., Han, T. S. & Morrison, C. E. (1995) Waist circumference as a measure for indicating need for weight management. Br. Med. J. 311: 158 –161.

4. Han, T. S., van Leer, E. M., Seidell, J. C. & Lean, M. E. (1996) Waist circumference as a screening tool for cardiovascular risk factors: evaluation of receiver operating characteristics (ROC). Obes. Res. 4: 533–547. 5. Han, T. S., van Leer, E. M., Seidell, J. C. & Lean, M. E. (1995) Waist circumference action levels in the identification of cardiovascular risk factors: prevalence study in a random sample. Br. Med. J. 311: 1401–1405. 6. National Institutes of Health (1998) Clinical Guidelines on the Identification, Evaluation, and Treatment of Overweight and Obesity in Adults: The Evidence Report. Obes. Res. 6 (suppl. 2): S51–S210. 7. Scottish Intercollegiate Guideline Network (SIGN) (1996) Obesity in Scotland. Integrating Prevention with Weight Management. A National Clinical Guideline Recommended for Use in Scotland by the Scottish Intercollegiate Guidelines Network. Pilot Edition. Edinburgh, Scotland. 8. Janssen, I., Katzmarzyk, P. T. & Ross, R. (2002) Body mass index, waist circumference, and health risk: evidence in support of current National Institutes of Health guidelines. Arch. Intern. Med. 162: 2074 –2079. 9. Bigaard, J., Tjønneland, A., Thomsen, B. L., Overvad, K., Heitmann, B. L. & Sørensen, T.I.A. (2003) Waist circumference, BMI, smoking, and mortality in middle-aged men and women. Obes. Res. 11: 895–903. 10. Kahn, H. S. & Valdez, R. (2003) Metabolic risks identified by the combination of enlarged waist and elevated triacylglycerol concentration. Am. J. Clin. Nutr. 78: 928 –934. 11. Zhang, X., Shu, X. O., Gao, Y. T., Yang, G., Matthews, C. E., Li, Q., Li, H., Jin, F. & Zheng, W. (2004) Anthropometric predictors of coronary heart disease in Chinese women. Int. J. Obes. Relat. Metab. Disord. 28: 734 – 40. 12. Moore, L. L., Bradlee, M. L., Singer, M. R., Splansky, G. L., Proctor, M. H., Ellison, R. C. & Kreger, B. E. (2004) BMI and waist circumference as predictors of lifetime colon cancer risk in Framingham Study adults. Int. J. Obes. Relat. Metab. Disord. 28: 559 –567. 13. Kushi, L. H., Kaye, S. A., Folsom, A. R., Soler, J. T. & Prineas, R. J. (1988) Accuracy and reliability of self-measurement of body girths. Am. J. Epidemiol. 128: 740 –748. 14. Hall, T. R. & Young, T. B. (1989) A validation study of body fat distribution as determined by self-measurement of waist and hip circumference. Int. J. Obes. 13: 801– 807. 15. Rimm, E. B., Stampfer, M. J., Colditz, G. A., Chute, C. G., Litin, L. B. & Willett, W. C. (1990) Validity of self-reported waist and hip circumferences in men and women. Epidemiology 1: 466 – 473. 16. Freudenheim, J. L. & Darrow, S. L. (1991) Accuracy of self-measurement of body fat distribution by waist, hip, and thigh circumferences. Nutr. Cancer 15: 179 –186. 17. Sonnenschein, E. G., Kim, M. Y., Pasternack, B. S. & Toniolo, P. G. (1993) Sources of variability in waist and hip measurements in middle-aged women. Am. J. Epidemiol. 138: 301–309. 18. Weaver, T. W., Kushi, L. H., McGovern, P. G., Potter, J. D., Rich, S. S., King, R. A., Whitbeck, J., Greenstein, J. & Sellers, T. A. (1996) Validation study of self-reported measures of fat distribution. Int. J. Obes. Relat. Metab. Disord. 20: 644 – 650. 19. Roberts, C. A., Wilder, L. B., Jackson, R. T., Moy, T. F. & Becker, D. M. (1997) Accuracy of self-measurement of waist and hip circumference in men and women. J. Am. Diet. Assoc. 97: 534 –536. 20. Han, T. S. & Lean, M. E. (1998) Self-reported waist circumference compared with the ’Waist Watcher’ tape-measure to identify individuals at increased health risk through intra-abdominal fat accumulation. Br. J. Nutr. 80: 81– 88. 21. Spencer, E. A., Roddam, A. W. & Key, T. J. (2004) Accuracy of self-reported waist and hip measurements in 4492 EPIC-Oxford participants. Public Health Nutr. 7: 723–727. 22. Bland, J. M. & Altman, D. G. (1986) Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1: 307–310. 23. Bland, J. M. & Altman, D. G. (2003) Applying the right statistics: analyses of measurement studies. Ultrasound Obstet. Gynecol. 22: 85–93. 24. Koh-Banerjee, P., Chu, N. F., Spiegelman, D., Rosner, B., Colditz, G., Willett, W. & Rimm, E. (2003) Prospective study of the association of changes in dietary intake, physical activity, alcohol consumption, and smoking with 9-y gain in waist circumference among 16 587 US men. Am. J. Clin. Nutr. 78: 719 –727. 25. Lawlor, D. A., Bedford, C., Taylor, M. & Ebrahim, S. (2002) Agreement between measured and self-reported weight in older women. Results from the British Women’s Heart and Health Study. Age Ageing 31: 169 –174. 26. Wang, J., Thornton, J. C., Bari, S., Williamson, B., Gallagher, D., Heymsfield, S. B., Horlick, M., Kotler, D., Laferrere, B., et al. (2003) Comparisons of waist circumferences measured at 4 sites. Am. J. Clin. Nutr. 77: 379 –384.

Downloaded from jn.nutrition.org by on April 18, 2009

lower rib, which was again smaller than the waist circumference measured halfway between the lower rib and the iliac crest, and that the waist circumference measured immediately above the iliac crest was largest (26). Waist circumference at the level of the umbilicus probably corresponds to a measurement immediately above the iliac crest. The difference between the waist circumference measurements at the 2 body sites depended on body shape (waist and hip circumference). Measuring the waist circumference at the level of the umbilicus includes an aspect of the hip circumference and this may explain the poorer agreement between the 2 waist circumference measurements among the women even though measured by technicians. The high correlation between self-reported waist circumference at the level of umbilicus and technician-measured waist circumference at the natural waist indicated that the self-reported waist circumference is valuable as a proxy for the technician-measured waist circumference in future studies of changes in waist circumference in the Danish Diet, Cancer and Health study. However, the mean change will be overestimated for women and slightly underestimated for men, which could lead to misclassification in categorical approaches. Perhaps more importantly, the association of the difference between the 2 measures with the baseline BMI (men) and baseline waist circumference (women) indicated that it is imperative that future analyses be adjusted for these variables because otherwise the measurement error may capture the association with baseline waist circumference and BMI, which would bias the results. In conclusion, self-reported measurements underestimated waist circumference in both men and women, and the underestimation increased with increasing body size. The limits of agreement were very wide, indicating poor agreement between measurements. Waist circumference measured at the level of the umbilicus was larger than that measured at the natural waist; the limits of agreement were acceptable from a clinical point of view in men, but very wide in women, indicating poor agreement. If possible, the same body sites for measurements of the waist circumference at baseline and at follow-up would be preferred in longitudinal studies. The self-reported waist circumference seemed to be usable as a proxy for technicianmeasured waist circumference in regression analyses, but such analyses should be adjusted for baseline waist circumference and baseline BMI.