Agreement between body mass index, waist circumference and skin ...

Annals of Human Biology, May–June 2011; 38(3): 257–264 Copyright q Informa UK, Ltd. ISSN 0301-4460 print/ISSN 1464-5033 online DOI: 10.3109/03014460.2010.517566

RESEARCH PAPER

Agreement between body mass index, waist circumference and skin-fold thickness in the United Kingdom Army Roberto J. Rona1, Josefin Sundin2, Paula Wood3 & Nicola T. Fear2 King’s College London, King’s Centre for Military Health Research, London, UK, 2King’s College London, Academic Centre for Defence Mental Health, Department of Psychological Medicine, London, UK, and 3Defence Science and Technology Laboratory (DSTL), Farnborough, UK

Ann Hum Biol Downloaded from informahealthcare.com by Kings College London on 08/04/11 For personal use only.

1

percentage of the US military considered overweight has increased in all age groups (Anonymous 2009), as has the percentage of obesity in US military (Bray et al. 2006). The meaning of such an increase based on body mass index (BMI) is not straight-forward. There has been a steep increase in body weight, with just a moderate increase in abdominal circumference in the US military between 1864 and 2000 (Friedl 2004). Thus, it is possible that an individual is large, but not obese. BMI is the usual indicator to assess overweight and obesity in population studies, but its validity has been questioned because the index does not measure fat only (Friedl 2004; Heinrich et al. 2008). In the military such lack of specificity may be a drawback because muscle mass may be a greater component of BMI than in the general population. This drawback may be greater in younger than in older service personnel as muscle mass decreases with age (Gallagher et al. 1996). In addition, physical performance in the military is an occupational requirement and fat-free mass, which is highly associated with BMI, is also highly associated with maximal performance (Fitzgerald et al. 1986; Harman and Frykman 1992). Although the military could be considered a special group it is not an uncommon occupation and the levels of fitness required may be a common feature of other occupations such as the police, security guards, excise and custom personnel, prison officers and physical education teachers. Hitherto overweight and obesity in the UK military have not been monitored. As part of the periodic health examination, service personnel are weighed and given advice. However, there is not a routine information system to estimate obesity in the UK Armed Forces. Although there is a document for assessment of medical fitness for military service based on BMI (Lunt et al. 2007), it is unclear how healthcare staff interpret height and weight measurements to advise and manage obesity. In the US military, age- and

Background: Body Mass Index (BMI) does not distinguish between fat-free mass and fat mass, yet this distinction is important as a clinical tool, especially in the military and occupations for which physical fitness is important. Aim: This study assessed the level of agreement between BMI, sum of skin-folds and waist circumference in UK Army personnel. Subjects and methods: Three hundred and eighty-six males were selected to study the level of agreement between BMI, waist circumference, waist circumference/height and sum of skin-folds at four sites. As the units of measurement differ, this study generated standardized distributions with skewness of zero. Results: The limits of agreement between BMI and waist circumference were fairly wide (between boundaries 1.75 z-score) and wide between BMI and sum of skin-folds (2.57 z-score) and waist circumference and skin-folds (2.37 z-score). The standard deviations of the differences of BMI and waist circumference decreased with increasing means of these two measures. The kappa statistic for BMI and waist circumference was 0.7 and , 0.5 if compared against sum of skin-folds. Conclusion: The level of agreement between BMI and waist circumference is acceptable regardless of method to assess obesity, but would be poor for assessing thinness. The level of agreement with sum of skin-folds was mediocre. Keywords: Kappa statistics, limits of agreement, military, obesity, public health

INTRODUCTION Preventing overweight and obesity is essential in the military, not only to decrease the risk of cardiovascular diseases and other chronic conditions (Mokdad et al. 2003; Ogden et al. 2007), but also to increase fitness and operational preparedness. Over the last 10 years the annual

Correspondence: Professor Roberto J. Rona, King’s Centre for Military Health Research, Kingston Education Centre, Weston Education Centre, Cutcombe Rd, London SE5 9RJ, UK. Tel: þ 44 207 848 5049. Fax: þ 44 207 848 5408. E-mail: [email protected] (Received 17 May 2010; accepted 15 August 2010)

257


258

R. J. RONA ET AL.

gender-specific weight-for-height allowances are used as the initial screening measure. Waist circumference is assessed as part of a gender-specific function to estimate body fat in those who have exceeded their weight allowance (Hodgdon and Friedl 1999; Friedl 2004; Williamson et al. 2009). The paradigm that BMI and waist circumference should be used when assessing obesity is backed by this US research. However, we do not know the degree of agreement between these two measures. This is important because not only BMI has limitations. Although waist circumference has been gaining popularity alone or as waist circumference/height ratio (Han et al. 1995; McCarthy and Ashwell 2006), it is a regional measure which does not necessarily reflect total fatness. Furthermore, the level of disagreement between BMI and waist circumference may vary systematically across the range of values. The assessment of agreement has been used to compare two methods based on continuous clinical measurements when neither provides an unequivocally correct measurement (Bland and Altman 1986). It has the advantage that it examines the agreement between two measurements along the entire distribution rather than making an assessment based on categories of truncated data. If the level of agreement between BMI and waist circumference were high the use of these two measures in clinical settings would be of limited value. On the other hand, if the level of disagreement were high the use of these two measures would be appropriate. It would be also meaningful to explore the most appropriate manner to use these two measurements in the Medical Centres. For example, using a two-stage approach, first measuring BMI in all members of the Armed Forces during their periodic examination and waist circumference only in those who are above a pre-defined BMI level, say 25 or over. There is a need to clarify the situation as 24% of males and females in the English general population had a BMI of 30 or more, denoting obesity (Zaninotto et al. 2009). The mean BMI in 20 – 24 year old recruits in the UK has increased from 22 to 23 from 1900 to 1980, reaching a mean of , 25 by 2007 (Rosenbaum and Crowdy 1992; Rona 1998; Wood 2007). Although the use of BMI and waist circumference in clinical settings is more feasible than skin-fold thickness, the sum of skin-folds in four sites has been considered a good measure of total body fat (Durnin and Womersley 1974). It has not enjoyed popularity in adult surveys as it is usually considered technically challenging. However, in the hands of well trained fieldworkers it could be considered a helpful measurement to assess fatness, especially when dual energy x-ray absorptiometry (DXA) or another gold standard is unavailable (Van der Ploeg et al. 2003). Contrary to common belief it is not known if BMI is a poorer health measure than waist circumference. Although waist circumference has been preferred over BMI as an assessment of obesity, both perform equally well in their association with known cardio-vascular risk factors such as blood pressure and insulin-mediated glucose update (Farin et al. 2005; Huxley et al. 2010).

In 2007, height, weight, waist circumference and sum of skin-folds in four sites were measured in 386 male Army personnel. Based on this study, we have assessed the level of agreement using the Bland-Altman method for continuous data (Bland and Altman 1986) and the kappa statistic for categorical data, between BMI, waist circumference and the sum of skin-folds. Assessment of agreement using continuous variables between these measures has not been carried out so far. Our aim was to assess the level of agreement between the three measures of obesity, BMI, waist circumference and sum of skin-folds. From a policy point of view, we wanted to appraise whether BMI on its own would be sufficient to assess obesity in the UK Armed Forces. This aspect needs to be clarified because waist circumference is not usually measured in Military Medical Centres. We believe that our results would be applicable to other occupations which have a predominantly young population for whom physical fitness is imperative. METHODS A sample of 386 male Army personnel was recruited for this study from units in the UK and Germany as identified by the Defence Science and Technology Laboratory (Dstl). The units invited to participate had an excess of combat personnel in comparison to the Army as a whole. Women were excluded because of their small number (n ¼ 28) in this study. In this study height was measured with the Leicester stadiometer. Participants were instructed to stand up straight with eyes looking horizontal and the vertex of the skull in the centre of the upper measuring platform. Height was recorded to the nearest 1 millimetre; weight was measured with calibrated SECA scales to the nearest 200 grams. Participants were asked to place both feet on the scales and look forward to the horizon. Subjects were measured wearing combat trousers, but objects were removed from pockets. Recorded weights were adjusted for the weight of the garments worn (1.2 kilograms). Waist circumference was measured, at the midway between lower rib and iliac crest with the tape measure horizontal. Subjects were instructed to stand with arms by the side and the measurement was taken to the nearest 1 millimetre after exhaling; skin-fold thickness measurements were taken with a Harpenden instrument on the right side of the body. The sites for triceps and biceps skin-fold measurements were located with the arm bent at the elbow at right angles with the palm up. The calliper was applied midway between the acromial and the olecranon processes. A fold containing skin and subcutaneous tissue was gently pulled away and the callipers placed onto the fold. The subscapular measurement was taken holding a diagonal fold 1 – 2 centimetres below the inferior angle of the scapula. The suprailiac measurement was taken from a diagonal fold in the anterior axillary line, superior to and in line with the natural angle of the iliac. All measurements were taken 2 – 3 seconds after applying the callipers. Annals of Human Biology

AGREEMENT BETWEEN MEASURES OF FATNESS IN THE UK ARMY


All measurements were taken by one of the authors (PW), except that a trained technician measured skin-fold thickness in 14 individuals. PW and the technician carried out an assessment in 18 subjects before the study started which demonstrated high repeatability. The average of the two measurements was used in the analysis. Information was also collected on age, geographical location of barrack, rank and, for 232 (61%) of the participants, role (combat, combat support and combat service support). Statistical analysis The main analysis is based on the method described by Bland and Altman (1986). They proposed that the most informative way to compare two techniques based on continuous measurements is a plot of the differences between the two techniques (vertical axis) against the mean of the two techniques (horizontal axis). The mean difference would provide the level of bias between the two techniques and the standard deviation (SD) of the differences would be used to estimate the limits of agreement as the mean difference minus twice the SD and the mean difference plus twice the SD. The method described by Bland and Altman (1986) to assess agreement between two measurements uses the same units, but because the units of measurement for BMI (Kg/m2), waist circumference (cm) and sum of skinfolds (mm of thickness for 40 mm of surface at a constant pressure for mm2) are different we generated distributions with a skewness of zero based on logarithmic functions for BMI, waist circumference, waist circumference/height ratio and sum of skin-folds. As the four anthropometric measurements were not normally distributed they were log transformed to generate the function ln (exp-k), where k and the sign of exp is selected to a skewness equal to zero. The new variables were normal standardized resulting in z-scores with a mean of 0 and deviation of 1. These z-scores were used for estimating limits of agreement and obtaining plots. For the purpose of this paper we consider wide limits of agreement 2 z-scores or more, between 1.5 – 1.99 z-scores as fairly wide and less than 1.5 z-scores as reasonable agreement. We did not include correlations in the analysis as they have been shown to be inappropriate for assessing agreement between the two methods (Bland and Altman 1990). We also assessed the variation in the SD of the differences between the two measurements by quintiles of the distribution of the mean of the two measurements to assess whether the limits of agreement vary across the range of values (Rothwell 2000). We also assessed agreement between the measures used as categorical variables. BMI was categorized as normal weight (up to 24.9), overweight (25 – 29.9) and obese (30 and above). The other three measurements were each divided into three categories according to the percentage distribution of the three BMI categories (Table I). This was an appropriate procedure to ensure that the categories would be comparable between the measurements. This provided us with tables which allowed us to calculate Kappa statistics (Fleiss 1986). We assessed the level of agreement and q Informa UK, Ltd.

259

Table I. Percentages and inter-quartile ranges (IQR) of military personnel distribution for waist circumference, waist circumference/ height and sum of skin-folds following the BMI distribution for normal (24.9 or less), overweight (25.0– 29.9) and obese (30 or over) distribution. Variables and levels (IQR)

n (%)

Waist circumference Low (IQR in cm 77.6– 82.6) Mid (IQR in cm 88.1– 95.9) High (IQR in cm 103.3 –111.6)

188 (48.7) 155 (40.2) 43 (11.1)

Waist circumference/height Low (IQR in cm/cm 0.044 – 0.047) Mid (IQR in cm/cm 0.049 – 0.054) High (IQR in cm/cm 0.059 – 0.064)

188 (48.7) 154 (39.9) 44 (11.4)

Sum of skin-folds Low (IQR in mm 24.5– 32.8) Mid (IQR in mm 41.3 – 55.8) High (IQR in mm 73.5 –93.1)

190 (49.2) 153 (39.6) 43 (11.1)

whether the lack of agreement was uniformly distributed above and below the diagonal line of agreement. All analyses were carried out in Stata 10 (Stata Corporation, College Station, TX). The study was approved by the Ministry of Defence Research Ethics Committee. RESULTS The mean age of the studied personnel was 27.9 years (Table II). The majority of participants were noncommissioned officers and lower ranks. The mean BMI was 25.7 units, the mean waist circumference was 87.9 cm, mean waist circumference/height was 0.5 and mean sum of skin-folds in four sites was 43.1 mm (Table II). Figure 1 gives the Bland and Altman plots of agreement for BMI and the three other measurements. In the three graphs the mean differences were near zero, as all the distribution were z-standardised. The limits of agreement in terms of z-scores were fairly wide for BMI and waist circumference (0.84, 2 0.91, that is 1.75 z-score between upper and lower limits), BMI and waist circumference/ height (0.93, 2 0.90, distance between boundaries 1.83 z-score) and wide for BMI and sum of skin-folds (1.37, 2 1.20, distance between boundaries 2.57 z-score). The limits of agreement for sum of skin-folds and waist circumference were wide (1.31, 2 1.06, distance between boundaries 2.37 z-score), but slightly less than for BMI and sum of skin-folds (Figure 2). Only the limits of agreement between waist circumference and waist circumference/ height were reasonably narrow (0.76, 2 0.66 distance between boundaries 1.42 z-scores) (Figure 2), but the two measurements are not independent, as waist circumference is common to them both. Figure 3 gives the SD of the difference and bias dividing the sample in quintiles according to the mean of the two measures using the same axes as in Figure 1.The three graphs show a decrease of the SD as the mean of the two measures increases. This decrease was smooth for the mean of BMI and waist circumference and less so for the other two slopes. Thus, the level of agreement increases as the mean of the two

260

R. J. RONA ET AL.

Table II. General characteristics of sample (n ¼ 386). The sample includes males who have complete data on the height and weight measures.


Variable Location UK Germany Northern Ireland Other Partake in sport* Service Arm** Combat Combat Support Combat Support Services Rank Other Non-Commissioned Officer Officer Medically downgraded*** BMI (Kg/m 2) Normal range (24.9 or less) Overweight (25 –29.9) Obese (30 or over) Mid-point Waist Circumference (cm) Waist Circumference (cm) Normal range (below 94) High risk (94 –101.9) Obese (102 or over) Continuous variables Age BMI Waist circumference/height (cm/cm) Sum of skin-folds (mm)

n (%)

M (SD)

186 (48.6) 185 (48.3) 5 (1.3) 7 (1.8) 95 (24.6)

DISCUSSION The limits of agreement based on continuous variables using the Bland and Altman method between waist circumference and waist circumference/height was moderate, between BMI and waist circumference was fairly wide and waist circumference and sum of skin-folds and between BMI and sum of skin-folds were wide. However, the SD of the difference of two measurements decreased with increasing mean of the two measurements in all the comparisons. The kappa analyses showed that the level of agreement was good in relation to BMI and waist circumference, but only moderate in relation to any measurement compared to sum of skin-folds. We have no evidence that a larger than expected proportion of personnel with a high BMI had normal waist circumference, as lack of agreement was symmetrical.

138 (58.0) 79 (33.2) 21 (8.8) 185 189 12 49

circumference implying high risk of obesity. The majority of those with a BMI of 30 or more had a waist circumference of 102 cm or more. Although the majority of those with a BMI between 25 – 29.9 had a waist circumference implying normal waist circumference, a sizeable number were at high risk of obesity or were obese.

(47.9) (49.0) (3.1) (12.7)

189 (49.0) 154 (39.9) 43 (11.1) 87.9 (10.2) 293 (75.9) 53 (13.7) 40 (10.4) 27.9 (5.9) 25.7 (3.8) 0.5 (0.1) 43.1 (20.0)

*Response to a single question on taking part in sport activities; **Only data on n ¼ 238; ***Level of employability, if downgraded for medical reasons the level activity for an individual may be restricted according to the severity of the problem.

measurements increases. The same was observed for mean waist circumferences, with waist circumference and the sum of skin-folds (Figure 4). Biases did not follow any recognizable pattern (Figures 3 and 4). The kappa statistics were calculated following the BMI distribution shown in Table I. The kappa statistics with three categories were reasonably good between BMI and waist circumference (kappa 0.70, 95% CI 0.62– 0.78), BMI and waist circumference/height (kappa 0.68, 95% CI 0.60 – 0.76) and waist circumference and waist circumference/height (kappa 0.71, 95% CI 0.63 – 0.79). However, agreements which included sum of skin-folds were only moderate with any of the other measures: sum of skin-folds and BMI (0.47, 95% CI 0.39 –0.55) and sum of skin-folds and waist circumference (0.52, 95% CI 0.44 – 0.60). A common feature in these analyses was that the percentage of subjects in which one measure exceeded the second was similar to the percentage of subjects in which the second measure exceeded the first measure (results not shown). As the level of agreement between BMI and waist circumference varied according to the standardized mean of BMI and waist circumference, we carried out a crosstabulation between the two measures according to the usual thresholds for each measure in clinical practice (Table III). Only one individual with a BMI of less than 25 had a waist

Interpretation of results In the Armed Forces it is essential to distinguish between an increase in fat-free mass and an increase in fatness. BMI does not distinguish between the two. Thus, it is important to confirm whether those who have a BMI of 25 or more do have an increased level of fatness. Standard practice in the US is to determine obesity with BMI and, if a threshold is reached, calculations for circumference-based body fat measurements are made, based on waist circumference, neck circumference and height (Friedl 2004; Williamson et al. 2009). This approach has been criticized as it may overestimate the percentage of overweight in comparison to skin-fold measurements (Babcock et al. 2006). However, agreement between skin-fold measures and current US practice of assessing weight-for-height and waist circumference together with other measurements has not been appropriately assessed (Friedl and Bathalon 2006). We did not estimate percentage fatness based on previous validation studies, as available equations vary according to population and the gold standard used for validation (Durnin and Womersley 1974; Ward et al. 1975; Gallagher et al. 2000). A possible option to assess percentage of fatness would have been to use the equations developed for the US military (Harman and Frykman 1992; Friedl 2004), but such an approach would have been problematic because the measurements do not correspond to those used in our study. We would have liked to replicate the US findings on the validity of anthropometric data using UK military data before recommending its use. We focused our study on the agreement between measurements using z-scores which does not pre-judge whether one measurement is better than the other. An approach which is safe, taking into account that a recent Annals of Human Biology

AGREEMENT BETWEEN MEASURES OF FATNESS IN THE UK ARMY (b)

2

BMI (z-score) - waist circumference/height (z-score)

BMI (z-score) - waist circumference (z-score)

(a)

1

0

–1

–2 –2

–1

0

1

2

Mean difference

0

–1

–2 –2

–1

Mean ± 1.96SD

(c) BMI (z-score) - skinfold thickness (z-score)

1

0

1

2

3

Mean of BMI (z-score) and waist circumference/height (z-score) Mean difference

Mean difference –0.04, limits of agreement 0.84, –0.91.

Mean ± 1.96SD

Mean difference 0.02, limits of agreement 0.93, –0.90.

2 1 0 –1 –2 –2

–1

0

1

2

3

Mean of BMI (z-score) and skinfold thickness (z-score) Mean difference

Mean ± 1.96SD

Mean difference 0.09, limits of agreement 1.37, –1.20.

Figure 1. Bland and Altman plots of agreement of BMI and each of the other three measures to assess fatness; (a) waist circumference, (b) waist circumference/height and (c) sum of skin-folds.

The fact that the limits of agreement were wider when sum of skin-folds was one of the measurements could have two interpretations: that BMI and waist circumference may be only satisfactory as measures of obesity. Alternatively, skin-fold thickness in four sites is not a good measure to predict obesity. The first interpretation seems plausible, as Flegal et al. (2009) demonstrated that BMI and waist circumference may be an inaccurate measure of percentage of body fat. The analyses based on kappa statistics and on continuous measurements seemed inconsistent, by demonstrating that agreement between BMI and waist circumference based on broad categories is satisfactory in contrast to the analysis

study on personnel who had exceeded weight allowance in the US demonstrated that both waist circumference and BMI had a large percentage of false negatives when bioimpedance was used as gold standard (Heinrich et al. 2008). However, bioimpedance is not generally recognized as a gold standard measurement and the sample was based on a US military population already engaged in a prevention programme, as acknowledged by the researchers (Heinrich et al. 2008). Our findings indicate that the limits of agreement between the continuous measurements in our study are fairly wide. Thus, BMI cannot be readily exchanged with the other measurements to assess obesity in the military.

(b) Waist circumference (z-score) - skinfold thickness (z-score)

(a) Waist circumference (z-score) - waist circumference/height (z-score)


2

3

Mean of BMI (z-score) and waist circumference (z-score)

261

2 1 0 –1 –2 –2

–1

0

1

2

Mean of waist circumference (z-score) and waist circumference/height (z-score) Mean difference

Mean ± 1.96SD

3

2 1 0 –1 –2 –2

–1

0

1

2

3

Mean of waist circumference (z-score) and skinfold thickness (z-score) Mean difference

Mean ± 1.96SD

Figure 2. Bland and Altman plots of agreement between (a) waist circumference and waist circumference/height and (b) sum of skin-fold thickness and waist circumference. q Informa UK, Ltd.

R. J. RONA ET AL.

BMI (z-score) - waist circumference (z-score)

(a )

(b)

0.6

0.6 BMI (z-score) - waist circumference (z-score)

262

0.4

0.2

0

0.4

0.2

–0.2

0

–2.24 to –0.80 –0.80 to –0.23

–0.21 to 0.28

0.28 to 0.84

0.86 to 3.22

–2.40 to –0.82 –0.82 to –0.22 –0.22 to 0.20

Mean of BMI (z-score) and waist circumference (z-score) Bias

Standard deviation

Bias

0.83 to 3.04

Standard deviation

BMI (z-score) - skinfold thickness (z-score)

0.8

0.6

0.4

0.2

0 –2.43 to –0.82 –0.82 to –0.26

–0.25 to 0.17

0.17 to 0.81

0.82 to 3.07

Mean of BMI (z-score) and skinfold thickness (z-score) Bias

Standard deviation

Figure 3. Standard deviation (SD) and bias in quintiles using the Bland and Altman axes of agreement for BMI and (a) waist circumference; (b) waist circumference/height and (c) sum of skin-folds. (b) Waist circumference (z-score) - skinfold thickness (z-score)

(a ) Waist circumference (z-score) - waist circumference/height (z-score)


( c)

0.21 to 0.83

Mean of BMI (z-score) and waist circumference/height (z-score)

0.5 0.4 0.3 0.2 0.1 0 –2.47 to –0.80 –0.79 to –0.24 –0.24 to 0.44

0.44 to 0.88

0.89 to 2.87

Mean of waist circumference (z-score) and waist circumference/height (z-score) Bias

Standard deviation

0.8

0.6

0.4

0.2

0 –2.24 to –0.78 –0.77 to –0.33 –0.32 to 0.16

0.19 to 0.84

0.86 to 2.91

Mean of waist circumference (z-score) and skinfold thickness (z-score) Bias

Standard deviation

Figure 4. Standard deviation (SD) and bias of the difference between two measurements in quintiles using the Bland and Altman axes of agreement; (a) waist circumference and waist circumference/height and (b) waist circumference and sum of skin-folds. Table III. Cross-tabulation of BMI and waist circumference according to the usual cut-offs points in clinical practice. Waist circumference (cm) 2

BMI (Kg/m ) Normal range (# 24.9) Overweight (25– 29.9) Obese (30þ )

Normal range (, 94 cm) n (%)

High risk (94–101.9 cm) n (%)

Obese ($ 102 cm) n (%)

188 (99.5) 105 (68.2) 0 (0.0)

1 (0.5) 43 (27.9) 9 (20.9)

0 (0.0) 6 (3.9) 34 (79.1)

based on the method of Bland and Altman (1986). However, this apparent contradiction was explained in our study by demonstrating that the SD of the mean difference between BMI and waist circumference across the range of these two measurements decreased substantially as the mean increased. Thus, the level of agreement would be better in the region used to assess obesity and worse in the region to assess thinness. As research has focused more on assessing obesity our finding is also consistent with the conclusion that both BMI and waist circumference are closely related

(Heinrich et al. 2008; Flegal et al. 2009). However, agreement between these measures would be greater at levels of overt obesity than at moderate levels of obesity. Although it was not the main purpose of the study, our findings indicate that the UK military should pay attention to the way obesity is verified. In our sample, 11% of the group was obese based on the BMI categories used to assess obesity in the general population. This percentage, although lower than the current prevalence of obesity in the general male English population aged 16 –34 years, is still quite high Annals of Human Biology

AGREEMENT BETWEEN MEASURES OF FATNESS IN THE UK ARMY


(Zaninotto et al. 2009). The percentages for waist circumference of 94 cm or over and waist circumference/ height above 0.5 were equivalent to the general population (McCarthy and Ashwell 2006). Strengths and weaknesses The advantages of our study are that a wide range of measurements were taken and the great majority of them were taken by one professional, thus decreasing measurement error. We are not aware of another study which has assessed the level of agreement between measures of obesity, except an assessment of kappa coefficients using binary variables based on the Third National Health and Nutrition Examination Survey in 12 –18 year olds (Himes 1999). Prevalence rates of obesity and overweight from this study should be viewed with caution as the sample studied is not representative of the UK Armed Forces, as it was not the purpose of this study. The sample in our study was formed of young service personnel with a disproportionately high percentage of combat role personnel in comparison to the UK Armed Forces. We suspect that the true percentage of obesity could be higher in a representative sample of the UK Armed Forces. Obesity may be higher in those fulfilling administrative roles, combat service support roles such as catering and healthcare and in Royal Navy personnel who are constrained to ships with limited space for long periods. The percentage of commissioned officers in the study was low and our results may not apply to this group. DXA, densitometry or other gold standard measures of obesity were not included in this study, thus we do not make assumptions about the validity of the anthropometric measurements to identify degrees of fatness. Implications The indication from our study is that obesity could be a serious problem in the UK Armed Forces. There is a belief that BMI as the sole measure to assess obesity may be too limited. We would recommend that validation of anthropometric measurements suitable to the Armed Forces is undertaken. In the meantime, it is important to appraise whether BMI on its own or a two-stage assessment of obesity using BMI and, if necessary, waist circumference would be more appropriate. Our results indicate that BMI on its own would perform fairly well to diagnose obesity (BMI 30 or more), as the majority (80%) had a waist circumference of 102 cm or over and the rest were at high risk of obesity (94 – 101.9 cm). However, a BMI between 25 – 29.9 would be uninformative, as 68% in our study had a waist circumference below 94 cm, but a third were at high risk of obesity or obese. Thus, we believe that a two-stage assessment would be helpful, especially for those with a BMI of 25 – 29.9. In the first stage only BMI would be assessed, revealing those with a BMI of 25 or more, a threshold which would have high sensitivity but low specificity as would include many with high fat-free mass only. In the second stage waist circumference would be measured in those with a BMI of 25 or more. Those with a waist circumference of 102 cm or over would be deemed obese and in need of q Informa UK, Ltd.

263

appropriate management and follow-up. Those with a waist circumference between 94 – 101.9 would be considered at high risk of obesity and should be monitored and given advice on the risks of obesity and its prevention. This approach would greatly increase specificity and only slightly decrease sensitivity, as in our study only one individual among those with a BMI below 25 had a waist circumference denoting high risk of obesity. This approach would have identified 10.3% of individuals as obese and 13.5% at high risk of obesity in this study. Although the limits of agreement between BMI and waist circumference were fairly wide across the range of continuous measurement, it was reasonable in the higher levels of the continuum. We recommend a two-stage assessment of obesity using BMI in all staff personnel and measuring waist circumference in those above a BMI threshold of 25. ACKNOWLEDGEMENTS We thank Lt Col Alan Ellis (HSG) Military for facilitating the recruitment of volunteers for the study and Andrew Amor and Venturino R. Nevola, who helped with the collation of the anthropometric data. The study was funded by the Ministry of Defence. Declaration of interest: Mrs Paula Wood works in Dstl which is part of the Ministry of Defence. The authors alone are responsible for the content and writing of the paper. REFERENCES Anonymous. 2009. Diagnosis of overweight/obesity, active component, US Armed Forces 1998 – 2008. MSMR (Medical Surveillance Monthly Report) 16(01):2– 7. Available online at: http://afhsc. army.mil/msmr_pdfs/2009/v16_n01.pdf#Page¼ 01, accessed 17 May 2009. Babcock CJ, Kirby TE, McCarroll ML, Devor ST. 2006. A comparison of military circumference equations to skinfold-based equations to estimate body composition. Mil Med 171:60– 63. Bland JM, Altman DG. 1986. Statistical methods of clinical measurement. Lancet 327:307 – 310. Bland JM, Altman DG. 1990. A note on the use of the intraclass correlation coefficient in the evaluation of agreement between two methods of measurement. Comput Biol Med 20:337 – 340. Bray RM, Hourani LL, Olmsted KL, Witt M, Brown JM, Pemberton MR, Marsden ME, Marriott B, Vandermaas-Peeler R, Beimar B, Calvin S, Bradshaw M, Close K, Hayden D. 2006. Department of Defense survey health related behaviors among active duty personnel. Available online at: http://www.ha.osd.mil/special_ reports/2005_Health_Behaviors_1-07.pdf, accessed 7 September 2010. Durnin JVGA, Womersley J. 1974. Body fat assessed from total body density and its estimation from skinfold thickness: Measurements on 481 men and women aged from 16 to 72 years. Br J Nut 32: 77–97. Farin HM, Abbasi F, Reaven GM. 2005. Body mass index and waist circumference correlate to the same degree with insulin-mediated glucose uptake. Metabolism 54:1323 – 1328. Fitzgerald PI, Vogel JA, Daniels WL, Dziados JE, Teves MA, Mello RP, Reich PJ. 1986. The body composition project: A summary report and description data. Technical Report No. T5-87. Natick, MA: US Army Research Institute of Environmental Medicine.


264

R. J. RONA ET AL.

Flegal KM, Shepherd JA, Looker AC, Graubard BI, Borrud LG, Ogden CL, Harris TB, Everhart JE, Schenker N. 2009. Comparisons of percentage body fat, body mass index, waist circumference, and waist-stature in adults. Am J Clin Nut 89:500 –508. Fleiss JL. 1986. The design and analysis of clinical experiments. New York: Wiley. Friedl KE. 2004. Can you be large and not obese? The distinction between, body weight, body fat, and abdominal fat in occupational standards. Diabetes Technol Ther 6:732 – 749. Friedl KE, Bathalon GP. 2006. Letters to the editor. Mil Med 171:vi. Gallagher D, Heymsfield SB, Heo M, Jebb SA, Murgatroyd PR, Sakamoto Y. 2000. Healthy percentage body fat ranges: An approach for developing guidelines based on body mass index. Am J Clin Nutr 72:694– 701. Gallagher D, Visser M, Sepulveda D, Pierson RN, Harris T, Heymsfield SB. 1996. How useful is body mass index for comparison of body mass index across age, sex, and ethnic groups? Am J Epidemiol 143:228 – 239. Han TS, van Leer EM, Seidell JC, Lean MEJ. 1995. Waist circumference action levels in the identification of cardiovascular risk factors: Prevalence in a random sample. BMJ 311:1401 –1405. Harman EA, Frykman PN. 1992. The relationship of body size and composition to the performance of physically demanding tasks. In: Marriott BM, Grumstrup-Scott J, editors. Body composition and physical performance: Applications for the military services. Washington, DC: National Academy Press. p 105– 118. Heinrich KM, Jitnarim N, Suminski RR, Berkel L, Hunter CM, Alvarez L, Brundige AR, Peterson AL, Foreyt JP, Haddock CK, Poston WSC. 2008. Obesity classification in military personnel: A comparison of body fat, waist circumference, and body mass index measurements. Military Med 173:67 –73. Himes JH. 1999. Agreement among anthropometric indicators identifying the fattest adolescents. Int J Obes 23 2(Suppl 2): S18– S21. Hodgdon JA, Friedl K. 1999. Development of the DoD body composition estimation equations. Technical Document No. 99-2B. San Diego: Naval Health Research Center. Huxley R, Mendis S, Zheleznyakov E, Reddy S, Chan J. 2010. Body mass index, waist circumference and waist:hip ratio as predictors of cardiovascular risk-a review of the literature. Eur J Clin Nutr 64: 16–22.

Lunt H, Fallowfield J, Allsop A. 2007. Review of the Metropolitan Life Insurance Company Height and Body Mass Actuarial Tables for assessment of medical fitness for military service as described in JSP 346. INM Report No. 2007.026 McCarthy HD, Ashwell M. 2006. A study of central fatness using waist-to-height ratios in UK children and adolescents over two decades support the simple message – “keep your waist circumference to less than half your height”. Int J Obesity 30: 988– 992. Mokdad AH, Ford ES, Bowman BA, Dietz WH, Vinicor F, Bales VS, Mark JS. 2003. Prevalence of obesity, diabetes, and obesity-related health risk factors, 2001. JAMA 289:76– 79. Ogden CL, Yanovski SZ, Carroll MD, Flegal KM. 2007. The epidemiology of obesity. Gastroenterology 132:2087– 2102. Rona RJ. 1998. Secular trend of stature and body mass index (BMI) in Britain in the 20th century. In: Bodzsar BE, Susanne C, editors. Secular growth changes in Europe. Budapest: Eotvos University Press. p 335 –349. Rosenbaum S, Crowdy JP. 1992. British Army recruits: 100 years of height and weights. JR Army Med Corps 138:81 –86. Rothwell PM. 2000. Analysis of agreement between measurements of continuous variables: General principle and lessons from studies of imaging of carotid stenosis. J Neurol 415:825 – 834. Van der Ploeg GE, Withers RT, Laforgia J. 2003. Percent body fat via DEXA: Comparison with a four compartment model. J Appl Physiol 94:499– 506. Ward GM, Krzywicki HJ, Rahman DP, Quaas RL, Nelson RA, Consolazio CF. 1975. Relationship of anthropometric measurements to body fat as determined by densitometry, potassium-40, and body water. Am J Clin Nut 28:162 – 169. Williamson DA, Bathalon GP, Sigrist LD, Allen HR, Friedl KE, Young AJ, Martin CK, Stewart TM, Burrell L, Han H, Hubbard Van S, Ryan D. 2009. Military services firness database: Development of a computerized physical fitness and weight management database for the US Army. Mil Med 174:1 – 8. Wood P. 2007. The prevalence of UK service personnel at risk of obesity-related ill health. DstlnTR27252 Draft. Dstl Farnborough, Hants: Human System Group. Zaninotto P, Head J, Stamatakis E, Wardle H, Mindell. 2009. Trends in obesity among adults in England from 1993 to 2004 by age and social class and projections of prevalence to 2012. J Epidemiol Community Health 63:140 –146.

Annals of Human Biology