Predictive Ability of Waist-to-Height in Relation to Adiposity in Children Is Not Improved With Age and Sex-Specific Values

nature publishing group Predictive Ability of Waist-to-Height in Relation to Adiposity in Children Is Not Improved With Age and Sex-Specific Values Rachael W. Taylor 1, Sheila M. Williams 2, Andrea M. Grant 1, Barry J. Taylor 3 and Ailsa Goulding 1 A waist-to-height ratio (WHtR) 0.5 indicates increased health risk in children and adults. However, because of residual correlation between WHtR and height in children, dividing waist circumference by height to the power of one may be insufficient to correctly adjust for height during growth. This study aimed to determine whether age and sex-specific exponents which properly adjust for height affect the predictive ability of WHtR to correctly discriminate between children with differing fat distribution. Total and regional body fat was measured by dual-energy X-ray absorptiometry (DXA) in 778 (49% male) children and adolescents. WHtR was calculated as waist/height 1 (WHtR a ), and using two published age and sex-specific exponents for height (WHtR b ) (1) (WHtR c ) (2), and compared with various DXA indexes of body composition using receiver operating curve analysis. 15% of males and 17% of females had a WHtR a 0.5, with corresponding figures of 8% and 27% for WHtR b, and 23% and 17% for WHtR c. WHtR a was significantly different from WHtR b (males only, P < 0.001) but not WHtR c (P = 0.121). Areas under the receiver operating curve (AUC) for WHtR a were significantly higher than AUCs for WHtR b or WHtR c in relation to DXA-measured body composition (AUCs 0.89 for WHtR a compared with AUCs of 0.71 0.84 for WHtR b and WHtR c ). Simply dividing waist circumference by height (WHtR a ) correctly discriminates between children and adolescents with low and high levels of total and central fat at least 90% of the time. Keeping your waist circumference to less than half your height provides an effective screening index of body composition during growth. Obesity (2011) 19, 1062 1068. doi:10.1038/oby.2010.217 Introduction Interest is growing in the efficacy of waist-to-height ratio (WHtR) as an index of cardiovascular risk (3 5) or body fat distribution (6) during growth. Some (7 9), but not all (4,10,11) studies in children suggest that WHtR provides a superior indicator of adverse metabolic profiles compared with BMI. Furthermore, unlike BMI, WHtR is only weakly correlated with age, negating the need to express values relative to age and sex. Thus it may be that the simple public health message keep your waist to less than half your height, is applicable in both children and adults (3,12). However, two recent papers (1,2) have questioned the statistical validity of the WHtR due to the residual correlation of WHtR with height in children. WHtR effectively raises height to the power of one, which means that it could over or under-adjust for the effect of height at different ages (1). Analyses from the National Health and Nutrition Examination Survey dataset demonstrates that the residual correlation with height ranges from 0.29 in males to 0.36 in females 2 18 years of age (1), with lower values in a smaller Australian dataset (2). Whether appropriately adjusting for height affects the relationship between WHtR and cardiovascular risk factors or body composition is not known. This aim of this study was to determine whether age and sex-specific exponents which properly adjust for height affect the predictive ability of WHtR to correctly discriminate between children with differing fat distributions. Methods and Procedures Data on total and regional body composition was available for 778 (n = 382 males) predominantly white children and adolescents aged 5 18 years who had participated in various studies investigating body composition and health in our laboratory from 1996 to 2007 (13 17). All studies were approved by the Lower South Ethics Committee and written informed consent was obtained from each participant or parent/guardian. A brief medical history was obtained by questionnaire and no participant was taking medication that would affect his or her body composition or had a history of constitutional delay in growth or maturation. Duplicate measures of height (wall-mounted stadiometer), 1 Department of Medical and Surgical Sciences, University of Otago, DSM, Dunedin, South Island, New Zealand; 2 Department of Preventive and Social Medicine, University of Otago, DSM, Dunedin, South Island, New Zealand; 3 Department of Women s and Children s Health, University of Otago, DSM, Dunedin, South Island, New Zealand. Correspondence: Rachael W. Taylor (rachael.taylor@otago.ac.nz) Received 7 February 2010; accepted 26 July 2010; published online 30 September 2010. doi:10.1038/oby.2010.217 1062 VOLUME 19 NUMBER 5 May 2011 www.obesityjournal.org

Table 1 Age and sex-specific exponents for each waist-to-height ratio (WHtR) Age WHtR a WHtR b WHtR c WHtR a WHtR b WHtR c 5 1.00 1.21 1.09 1.00 1.24 1.07 6 1.00 1.07 1.09 1.00 1.00 1.07 7 1.00 1.44 1.09 1.00 1.67 1.07 8 1.00 1.84 1.00 1.84 9 1.00 1.51 1.33 1.00 1.72 1.24 10 1.00 1.40 1.33 1.00 1.37 1.24 11 1.00 1.68 1.33 1.00 1.39 1.24 12 1.00 1.43 1.00 1.00 13 1.00 1.48 1.00 1.17 14 1.00 1.00 1.00 0.76 15 1.00 0.89 0.75 1.00 0.61 0.85 16 1.00 1.11 0.75 1.00 0.62 0.85 17 1.00 1.02 0.75 1.00 0.74 0.85 18 1.00 0.96 1.00 0.93 For WHtR a, height is adjusted by the power of 1, WHtR b has age and sex-specific power adjustment according to Tybor et al. (1) and WHtR c has age and sex-specific power adjustment according to Nambiar et al., (2) respectively (WHtR c restricted to 257 males and 267 females only because of age). weight (electronic scales), waist (minimum circumference between the rib cage and the iliac crest using a nonstretchable tape measure) and hip (maximum protuberance of the buttocks) circumferences were obtained with the participants wearing light clothing and no shoes using standard procedures (18). Measurements were obtained by several trained examiners following the same protocols. In our laboratory, coefficients of variation (CV%) for height, weight, and waist circumference in young children are <2%. BMI was calculated as weight in kilograms/ height in meters squared. Children were classified as overweight (BMI 85 94th) and obese ( 95th) according to change data capture reference data (19). WHtR was calculated as waist circumference (cm)/ height (cm) and presented in three ways: (i) according to standard definitions (3) and hereafter referred to as WHtR a, (ii) using the age and sex-specific constants developed by Tybor et al. (1) for children aged 5 18 years inclusive, hereafter referred to as WHtR b, and (iii) using the age and sex-specific constants developed by Nambiar et al. (2) for children aged 5 7, 9 11, and 15 17 years inclusive, hereafter referred to as WHtR c. Thus the WHtR c analyses are restricted to 524 children (67% of sample). The age- and sex-specific constants for WHtR a, WHtR b, and WHtR c are presented in Table 1. All dual-energy X-ray absorptiometry (DXA) measurements were performed and analyzed by one experienced operator with a Lunar DPX-L scanner (software package 4.7; Lunar, Madison, WI) using standard procedures. The scanner determines total fat mass (kg) and the fat content (kg) of specific anatomical regions including trunk and extremity fat (automatic default regions) and central and peripheral fat (manual regions of interest) as shown in Figure 1. The trunk region consists of the area bordered by a horizontal line below the chin, vertical borders lateral to the ribs and oblique lines passing through the femoral necks. The arm region consists of all tissues outside these lateral borders and the leg region all tissue below the oblique lines. A central region of interest (waist fat) was also determined using manual analysis, defined as the tissue within an upper horizontal border at the top of L1, vertical borders lateral to trunk soft tissue, and a lower border positioned superior to the iliac crests. In our laboratory the CVs for repeated in vivo scans on 10 adults were 2.6% for fat mass, 2.5% for fat percentage and, <3.5% for all regional measurements (20). Statistics were performed using STATA (10). Data are presented as means (s.d.) and log transformed if not normally distributed before a Figure 1 (a) Automatic (trunk, arm, and leg) and (b) manual (waist and hip) regions of interest. analysis. Sex-specific receiver operating characteristic curves were calculated as previously described (6,20). Briefly, the sensitivity (truepositive rate) and specificity (true-negative rate) of various cutoffs for the screening measures (WHtR a, WHtR b, and WHtR c ) were compared against the referent DXA measures: trunk fat mass (kg), trunk fat mass b obesity VOLUME 19 NUMBER 5 May 2011 1063

index (trunk fat mass/height 2 ) and waist fat mass (kg). Comparison with total fat mass (kg) and fat mass index (total fat mass/height 2 ) are also presented for interest. Positive (true-positive rate divided by false-positive rate) and negative (false-negative rate divided by true-negative rate) likelihood ratios (LR) are also presented (21). Because commonly accepted cutoffs for regional fat distribution do not exist in children, internal sexspecific z-scores were created for each referent measure. A z-score +1 (approximates the 85th percentile) indicated a subject as truly positive. The receiver operating curves were compared using the STATA procedure roccomp, which uses procedures appropriate for correlated variables, when comparing different WHtR indexes. Results Table 2 presents the characteristics of the study population. Although the mean BMI z-score of females was significantly higher than that of males (P = 0.016), the prevalence of overweight and obesity was similar in males (14 and 9%, respectively) and females (19 and 10%, P = 0.147). No sex differences were observed in WHtR a or WHtR c and a similar proportion of boys and girls had WHtR a values 0.5 (P = 0.171). However, slightly more boys had a WHtR c 0.5 than girls (P = 0.040). By Table 2 Characteristics of the study population n 382 396 Age (years) 11.2 (4.2) 11.5 (4.2) Height (m) 1.47 (0.24) 1.44 (0.21) Height z-score 0.21 (1.00) 0.17 (0.95) Weight (kg) 43.7 (19.7) 44.2 (19.3) Weight z-score 0.40 (0.92) 0.50 (0.93) BMI (kg/m 2 ) 19.2 (3.6) 20.1 (4.4) a BMI z-score 0.40 (0.89) 0.55 (0.87) a Lean mass (kg) 32.9 (14.5) 28.6 (9.8) a Fat mass (kg) 8.5 (97.2) 13.2 (10.0) a Fat mass index (FMI) 3.7 (2.6) 5.7 (3.4) a Percentage fat 18.3 (8.7) 26.8 (10.0) a Trunk fat mass (kg) 3.4 (3.5) 5.5 (4.7) a Trunk fat mass index (TFMI) 1.4 (1.3) 2.3 (1.7) a Waist fat (kg) 0.63 (0.67) 0.98 (0.80) a Waist circumference (cm) 67.0 (11.9) 65.8 (11.2) WHtR a 0.46 (0.05) 0.46 (0.05) WHtR a 0.5 (%) 15 17 WHtR b 0.42 (0.06) 0.46 (0.08) a WHtR a 0.5 (%) 8 27 WHtR c 0.47 (0.06) 0.46 (0.05) WHtR a 0.5 (%) 23 17 Data are presented as mean (s.d.) except where specified. FMI is calculated as fat mass divided by height in meters squared. TFMI is trunk fat divided by height in meters squared. WHtR is waist circumference divided by height. For WHtR a, height is adjusted by the power of 1, WHtR b has age and sex-specific power adjustment according to Tybor et al. (1) and WHtR c has age and sex-specific power adjustment according to Nambiar et al., (2) respectively (WHtR c is restricted to 257 males and 267 females only because of age). a P < 0.05 compared with males by independent t-test (analyzed on logtransformed data where not normally distributed). contrast, females had a significantly higher WHtR b (Table 1), with 27% of girls having a value 0.5 compared with only 8% of boys (P < 0.001). Figure 2 displays the difference in the various age and sexspecific WHtR indexes in relation to age. For WHtR b, the power to raise height which eliminates any correlation with height, tends to be greater than one in younger children (<14 years of age), resulting in lower values for WHtR b (1). In adolescent females, the exponent tends to be <1, with concomitant higher values for WHtR b. WHtR a and WHtR b were moderately correlated in the whole sample (r = 0.63, P < 0.001) and in both boys (r = 0.70, P < 0.001) and girls (r = 0.61, P < 0.001). Paired t-tests showed they were significantly different in boys and the whole sample (both P < 0.001), but not in girls (P = 0.309). However, the majority of children were classified by both WHtR indexes as below (n = 586, 75%) or above (n = 70, 9%) the 0.5 cutoff. The remaining 122 participants were split fairly evenly between those with low WHtR a and high WHtR b (n = 67) and high WHtR a and low WHtR b (n = 55). Differences between WHtR a and WHtR c were less pronounced than that between WHtR a and WHtR b, especially in girls (Figure 2). The correlation between WHtR a and WHtR c was higher in girls (r = 0.86, P < 0.001) than in boys (r = 0.67, P < 0.001), but no significant differences were observed in a WHtR difference (WHtR a WHtR b ) b WHtR difference (WHtR a WHtR c ) 0.2 0.15 0.1 0.05 0.05 0.1 0.15 0.2 0.05 0.05 0.1 0.15 0.2 WHtR a WHtR b 0 0 2 4 6 8 10 12 14 16 18 20 0.2 0.15 0.1 Age (years) WHtR a WHtR c 0 0 2 4 6 8 10 12 14 16 18 20 Age (years) Figure 2 Individual subject differences between the various waist-toheight ratio (WHtR) indexes in relation to age in males (filled circles) and females (open circles). (a) WHtR a WHtR b. (b) WHtR a WHtR c. 1064 VOLUME 19 NUMBER 5 May 2011 www.obesityjournal.org

Table 3 Predictive ability of each WHtR screening tool in relation to the reference measures WHtR a WHtR b WHtR c Sens Spec +LR LR Sens Spec +LR LR Sens Spec +LR LR Trunk fat mass 67 95 12.8 0.35 31 96 7.2 0.72 51 82 2.8 0.60 Trunk FMI 74 95 15.2 0.27 33 96 7.8 0.70 60 82 3.3 0.49 Waist fat mass 70 95 13.5 0.31 33 96 7.8 0.70 58 82 3.2 0.51 Total fat mass 64 94 11.6 0.38 32 96 8.0 0.71 53 82 2.9 0.58 Total FMI 69 95 14.0 0.33 31 96 7.2 0.63 58 82 3.2 0.52 Trunk fat mass 63 92 8.0 0.40 47 78 2.1 0.68 45 89 4.1 0.62 Trunk FMI 68 93 9.4 0.34 49 78 2.2 0.66 49 89 4.4 0.58 Waist fat mass 65 92 8.5 0.38 47 78 2.1 0.68 46 89 4.2 0.60 Total fat mass 62 92 8.1 0.41 52 79 2.5 0.61 49 90 4.8 0.57 Total FMI 71 93 10.2 0.31 57 80 2.8 0.54 55 90 5.5 0.50 Waist-to-height ratio (WHtR) is waist circumference divided by height. For WHtR a, height is adjusted by the power of 1, WHtR b has age and sex-specific power adjustment according to Tybor et al., (1) and WHtR c has age and sex-specific power adjustment according to Nambiar et al., (2) (WHtR c is restricted to 257 males and 267 females only because of age). Sens is the sensitivity or true-positive rate, spec is the specificity or true-negative rate, +LR is the positive likelihood ratio (true-positive rate divided by false-positive rate), LR is the negative likelihood ratio or falsenegative rate divided by the true-negative rate. FMI refers to fat mass index (fat divided by height squared). Table 4 Areas under the ROC curve (AUC) for each WHtR screening tool in relation to the reference measures WHtR a WHtR b WHtR c WHtR a WHtR b WHtR c Trunk fat mass 0.91 (0.86, 0.95) a 0.73 (0.66, 0.80) b 0.72 (0.62, 0.82) b 0.91 (0.87, 0.94) a 0.70 (0.63, 0.76) b 0.78 (0.70, 0.86) b Trunk fat mass index 0.94 (0.91, 0.97) a 0.73 (0.66, 0.81) b 0.76 (0.66, 0.86) b 0.93 (0.90, 0.96) a 0.70 (0.63, 0.77) b 0.80 (0.72, 0.88) c Waist fat mass 0.92 (0.88, 0.97) a 0.74 (0.66, 0.81) b 0.75 (0.65, 0.86) b 0.92 (0.88, 0.95) a 0.70 (0.63, 0.76) b 0.80 (0.72, 0.87) c Total fat mass 0.89 (0.85, 0.94) a 0.72 (0.64, 0.79) b 0.72 (0.62, 0.83) b 0.91 (0.87, 0.94) a 0.73 (0.66, 0.80) b 0.80 (0.73, 0.88) c Total fat mass index 0.93 (0.89, 0.96) a 0.71 (0.64, 0.78) b 0.73 (0.62, 0.83) b 0.93 (0.90, 0.96) a 0.74 (0.67, 0.81) b 0.84 (0.76, 0.91) c Data are presented as AUC (95% CI). Waist-to-height ratio (WHtR) is waist circumference divided by height. For WHtR 1, height is adjusted by the power of 1, WHtR 2 has age and sex-specific power adjustment according to Tybor et al., (1) and WHtR 3 has age and sex-specific power adjustment according to Nambiar et al., (2) respectively. Analysis for WHtR 3 is restricted to 257 males and 267 females only. ROC, receiver operating curve. a,b,c Different superscripts within the same line (separate for each sex) are significantly different (P < 0.05). either sex. Again, most children (87%) were classified similarly by WHtR a and WHtR c. Calculation of the κ-statistics demonstrated that agreement between the different indexes was moderate, with κ-values of 0.44 0.54 (P < 0.001). Regardless of the referent measure chosen, WHtR a showed only moderate sensitivity in both males (64 74%) and females (63 71%) with high DXA fat. By contrast, specificity was consistently higher in both sexes (92 95%). In males, WHtR b had similar high specificity but the sensitivity was considerably reduced, with only one-third of males with high DXA body fat being correctly classified by WHtR b. Classification of more centrally overweight females was also relatively poor, with only approximately half being correctly identified. Unlike in males, the specificity of WHtR b was also reduced in females (78 80%). The sensitivity and specificity of WHtR c was intermediate relative to the other screening tools in both sexes (Table 3). The LR presented in Table 3 further demonstrate the improved diagnostic ability of WHtR a relative to the other WHtR indexes; the positive LR was considerably higher and the negative LR markedly lower for WHtR a compared with WHtR b or WHtR c. Areas under the receiver operating curve (AUC) display the percentage of children for whom the screening tool correctly discriminates between those with low and high levels of body fat at all possible cutoffs for WHtR. For example, a randomly selected boy with high trunk fat mass would have a larger WHtR a than a randomly selected boy with low trunk fat mass 91% of the time (Table 4). Table 4 further illustrates that WHtR a correctly discriminates between different levels of body fat 89 94% of the time in males and 91 93% of the time in females. AUCs for WHtR b and WHtR c were similar in boys and both were significantly less likely to discriminate between boys according to their measured body fat regardless of the obesity VOLUME 19 NUMBER 5 May 2011 1065

referent measure chosen. Similarly, WHtR a was the superior measure in girls, although WHtR c performed significantly better than WHtR b. Discussion Our data demonstrate that the simple recommendation of having a waist less than half a person s height correctly discriminates between children and adolescents with low and high levels of central body fat more than 90% of the time, regardless of the DXA referent method chosen. It has been suggested that WHtR a is statistically flawed because it remains correlated with height in children (1,2). However, our study shows that use of age and sex-specific exponents to properly adjust for height, does not improve the predictive ability of WHtR to categorize children according to their DXA-measured fat distribution. In fact, for each DXA index chosen, simple WHtR a was significantly better than WHtR b or WHtR c at this task. The high specificity of WHtR a demonstrates that the vast majority of children who are not centrally obese, also have WHtR a values <0.5. However, as has been reported for BMI (22 24), WHtR a is a reasonably insensitive measure, correctly identifying only 63 74% of children who have high central body fat. We examined our data further to determine why the sensitivity of WHtR a is only moderate. The low WHtR a in those who were actually centrally obese (n = 40) arose from being tall for their age (mean z-score 1.2), rather than having a reduced waist circumference. These children had high trunk fat mass (mean z-score 1.4) above the cutoff denoting them as centrally obese, but were not as extreme as those children identified as having high WHtR a and high truncal fat (mean z-score 1.9). Conversely, children in the high WHtR a but low truncal fat group (n = 45) were so because they were short for their age (mean height z-score 0.32), rather than having excessively high waist circumferences. Overall, they had less body fat (partly related to their reduced height) and thus were not graded as centrally obese. An alternative explanation of the low sensitivity of WHtR in general concerns the choice of reference cutoff. Current accepted classifications for defining high central fat distribution or overall adiposity do not exist. Thus we defined a child as being truly positive if their age- and sex-adjusted DXAmeasured fat mass was above the 85th percentile within our study population. We chose the 85th percentile on the basis of current recommendations for defining overweight in children using BMI (25). However, if we used the 95th percentile to indicate high trunk fat mass, the sensitivity of WHtR a was substantially increased to 93% in boys and 91% in girls. This did not produce a corresponding decrease in specificity (92% and 87% in boys and girls, respectively). Correspondingly, the positive LR were high for both boys (11.7) and girls (6.9), with low negative LR (0.07 and 0.10 in boys and girls, respectively). Interestingly, WHtR b did not perform as well using the 95th percentile trunk fat cutoff, showing low sensitivity (53% in males and 59% in females) and moderate/high specificity (96% and 75%, respectively). As was observed throughout these analyses, the performance of WHtR c was in between the other two indexes (data not shown). Comparable data has been demonstrated with BMI during growth. Mei et al. (26) show BMI to be a very sensitive measure (>90%), but their percentage fat cutoffs (reference measure) were also much higher than is typically utilized in the literature (27,28). Similarly, Telford et al. (23) highlight the large discrepancy in reported sensitivity that can be produced with varying percentage body fat cutoffs used to denote overfat in children. In theory, use of the alternative exponents for height should remove any correlation between WHtR and height. In our sample, neither WHtR b (r = 0.12, P = 0.073) nor WHtR c (r = 0.03, P = 0.665) were correlated with height in boys (adjusted for age), but neither was WHtR a (r = 0.04, P = 0.607). By contrast, all indexes remained correlated with height in girls (WHtR a r = 0.18, P = 0.010: WHtR b r = 0.26, P < 0.001: WHtR c r = 0.28, P < 0.001). Further analysis showed significant correlations within some age groups in both sexes. However, despite our reasonable overall sample size, the sizes of many of these groups were possibly too small to be confident in these findings. This sexual dimorphism in the remaining relationship with height did not appear to affect the ability of WHtR to discriminate between those with and without high levels of central body fat, as evidenced by comparable AUCs for each index in males and females. BMI has also come under criticism for being residually correlated with height during growth (29 31). As Benn (32) highlighted many years ago, choosing the most appropriate index depends on two factors; first and most importantly, that it is highly correlated with adiposity and second, that the distribution of the index should ideally be independent of height. However, as was found in the present study for WHtR, studies have shown that alternative indexes of weight/height p are not necessarily superior indicators of either adiposity or cardiovascular risk during growth (33,34). Moreover, because height and body fat are related, it is feasible that exponents that minimize associations with height are unlikely to maximize the correlation with central fat mass. Furthermore, the performance of alternative weight/height p indexes is not consistent across the age spectrum for children (35). Thus, choosing the best index requires a pragmatic approach, as well as considering accuracy and statistical suitability. This is highlighted by examination of the proposed exponents for WHtR b from the National Health and Nutrition Examination Survey data (1). In general, they vary considerably from each year of age to the next. This means that the exponent that is unrelated to height changes from 1.44 in a boy aged 6.99 years to 1.84 in one aged 7.01 years, and thus changes the calculated WHtR by a considerable amount (up to 0.5 depending on the height and waist circumference of the child). This wide variability is observed even in studies based on very large datasets as used by Tybor et al. (1), and illustrates the impractical nature of determining different exponents for different age groups. As Nambiar et al. (2) conclude, although the error associated with WHtR a may be statistically significant, it is small (<2%) and thus is both clinically and biologically acceptable. Perhaps more importantly, WHtR a is a simple formula, which is straightforward 1066 VOLUME 19 NUMBER 5 May 2011 www.obesityjournal.org

and applies to most age groups and both sexes, without the need for varying exponents. Our study was conducted in a reasonably large database of predominantly Caucasian children over a large age range, with widely varying body composition. Our analyses are restricted to demonstrating the performance of WHtR in relation to measures of adiposity alone. As we did not have measures of cardiovascular risk we cannot comment on whether or not alternative WHtR indexes (WHtR b or WHtR c ) provide more optimal indicators of metabolic risk. There is also no commonly accepted method and cutoffs for defining high central body fat in children (20), so internal definitions such as ours are often used. Our waist circumference measurements were obtained at the minimum circumference of the abdomen, rather than at other common sites such as at the umbilicus, iliac crest, or midway between two bony landmarks (36). It is possible that we could have reached a different conclusion had we used an alternative waist measurement site. However, we consider this unlikely given that others have shown that waist circumference values measured at any of the four common sites are almost equally associated with trunk fat mass in adults of both sexes (36), and that site of measurement does not substantially influence the relationship between waist circumference and morbidity/mortality (37). Thus while WHtR a may be statistically flawed, alternative variations of WHtR offer no additional discriminatory power. WHtR a clearly showed the best combination of sensitivity and specificity for identifying children with low and high levels of regional body fat. Thus it is reassuring to know that the simple public health message keep your waist to less than half your height (3,12) provides a simple and pragmatic approach to discriminating between children who vary in regional body fat distribution. Acknowledgments R.W.T., A.G., and B.J.T were principal investigators of the various studies included. A.M.G. and A.G. undertook all DXA scans. S.M.W. completed all statistical analyses. R.W.T. wrote the first and subsequent versions of the manuscript, to which all authors contributed. Funding was received from the Health Research Council of New Zealand and the University of Otago, New Zealand. Disclosure The authors declared no conflict of interest. 2010 The Obesity Society REFERENCES 1. Tybor DJ, Lichtenstein AH, Dallal GE, Must A. Waist-to-height ratio is correlated with height in US children and adolescents aged 2 18 years. Int J Pediatr Obes 2008;3:148 151. 2. Nambiar S, Truby H, Abbott RA, Davies PS. Validating the waist-height ratio and developing centiles for use amongst children and adolescents. Acta Paediatr 2009;98:148 152. 3. Ashwell M, Hsieh SD. Six reasons why the waist-to-height ratio is a rapid and effective global indicator for health risks of obesity and how its use could simplify the international public health message on obesity. Int J Food Sci Nutr 2005;56:303 307. 4. Freedman DS, Kahn HS, Mei Z et al. Relation of body mass index and waist-to-height ratio to cardiovascular disease risk factors in children and adolescents: the Bogalusa Heart Study. Am J Clin Nutr 2007;86: 33 40. 5. Maffeis C, Banzato C, Talamini G; Obesity Study Group of the Italian Society of Pediatric Endocrinology and Diabetology. Waist-to-height ratio, a useful index to identify high metabolic risk in overweight children. J Pediatr 2008;152:207 213. 6. Taylor RW, Williams SM, Grant AM et al. Waist circumference as a measure of trunk fat mass in children aged 3 to 5 years. Int J Pediatr Obes 2008;3:226 233. 7. Savva SC, Tornaritis M, Savva ME et al. Waist circumference and waist-to-height ratio are better predictors of cardiovascular disease risk factors in children than body mass index. Int J Obes Relat Metab Disord 2000;24:1453 1458. 8. Hara M, Saitou E, Iwata F, Okada T, Harada K. Waist-to-height ratio is the best predictor of cardiovascular disease risk factors in Japanese schoolchildren. J Atheroscler Thromb 2002;9:127 132. 9. Kahn HS, Imperatore G, Cheng YJ. A population-based comparison of BMI percentiles and waist-to-height ratio for identifying cardiovascular risk in youth. J Pediatr 2005;146:482 488. 10. Mirzaei M, Taylor R, Morrell S, Leeder SR. Predictors of blood pressure in a cohort of school-aged children. Eur J Cardiovasc Prev Rehabil 2007;14:624 629. 11. Manios Y, Kourlaba G, Kafatos A et al. Associations of several anthropometric indices with insulin resistance in children: The Children Study. Acta Paediatr 2008;97:494 499. 12. McCarthy HD, Ashwell M. A study of central fatness using waist-to-height ratios in UK children and adolescents over two decades supports the simple message keep your waist circumference to less than half your height. Int J Obes 2006;30:988 992. 13. Goulding A, Cannan R, Williams SM et al. Bone mineral density in girls with forearm fractures. J Bone Miner Res 1998;13:143 148. 14. Goulding A, Jones IE, Taylor RW, Williams SM, Manning PJ. Bone mineral density and body composition in boys with distal forearm fractures: a dualenergy x-ray absorptiometry study. J Pediatr 2001;139:509 515. 15. Taylor RW, Goulding A, Lewis-Barned NJ, Williams SM. Rate of fat gain is faster in girls undergoing early adiposity rebound. Obes Res 2004;12: 1228 1230. 16. Miller JC, Grant AM, Drummond BF et al. DXA measurements confirm that parental perceptions of elevated adiposity in young children are poor. Obesity (Silver Spring) 2007;15:165 171. 17. Taylor RW, Murdoch L, Carter P et al. Longitudinal study of physical activity and inactivity in preschoolers: the FLAME study. Med Sci Sports Exerc 2009;41:96 102. 18. Parnell W, Scragg R, Wilson N, Schaaf D, Fitzgerald E. NZ food NZ children: key results of the 2002 National Children s Nutrition Survey. Ministry of Health: Wellington, New Zealand, 2003, pp 1 267. 19. Kuczmarski RJ, Ogden CL, Guo SS et al. 2000 CDC Growth Charts for the United States: methods and development. Vital Health Stat 11 2002;1 190. 20. Taylor RW, Jones IE, Williams SM, Goulding A. Evaluation of waist circumference, waist-to-hip ratio, and the conicity index as screening tools for high trunk fat mass, as measured by dual-energy X-ray absorptiometry, in children aged 3 19 y. Am J Clin Nutr 2000;72:490 495. 21. Biggerstaff BJ. Comparing diagnostic tests: a simple graphic using likelihood ratios. Stat Med 2000;19:649 663. 22. Sardinha LB, Going SB, Teixeira PJ, Lohman TG. Receiver operating characteristic analysis of body mass index, triceps skinfold thickness, and arm girth for obesity screening in children and adolescents. Am J Clin Nutr 1999;70:1090 1095. 23. Telford RD, Cunningham RB, Daly RM et al. Discordance of international adiposity classifications in Australian boys and girls - the LOOK study. Ann Hum Biol 2008;35:334 341. 24. Deurenberg-Yap M, Niti M, Foo LL, Ng SA, Loke KY. Diagnostic accuracy of anthropometric indices for obesity screening among Asian adolescents. Ann Acad Med Singap 2009;38:3 6. 25. Barlow SE; Expert Committee. Expert committee recommendations regarding the prevention, assessment, and treatment of child and adolescent overweight and obesity: summary report. Pediatrics 2007;120(Suppl 4):S164 S192. 26. Mei Z, Grummer-Strawn LM, Wang J et al. Do skinfold measurements provide additional information to body mass index in the assessment of body fatness among children and adolescents? Pediatrics 2007;119: e1306 e1313. 27. Taylor RW, Jones IE, Williams SM, Goulding A. Body fat percentages measured by dual-energy X-ray absorptiometry corresponding to recently obesity VOLUME 19 NUMBER 5 May 2011 1067

recommended body mass index cutoffs for overweight and obesity in children and adolescents aged 3 18 y. Am J Clin Nutr 2002;76:1416 1421. 28. McCarthy HD, Cole TJ, Fry T, Jebb SA, Prentice AM. Body fat reference curves for children. Int J Obes (Lond) 2006;30:598 602. 29. Cole TJ. Weight/heightp compared to weight/height2 for assessing adiposity in childhood: influence of age and bone age on p during puberty. Ann Hum Biol 1986;13:433 451. 30. Franklin MF. Comparison of weight and height relations in boys from 4 countries. Am J Clin Nutr 1999;70:157S 162S. 31. Hattori K, Hirohara T. Age change of power in weight/height(p) indices used as indicators of adiposity in Japanese. Am J Hum Biol 2002;14:275 279. 32. Benn RT. Some mathematical properties of weight-for-height indices used as measures of adiposity. Br J Prev Soc Med 1971;25:42 50. 33. Lazarus R, Baur L, Webb K, Blyth F. Adiposity and body mass indices in children: Benn s index and other weight for height indices as measures of relative adiposity. Int J Obes Relat Metab Disord 1996;20:406 412. 34. Frontini MG, Bao W, Elkasabany A, Srinivasan SR, Berenson G. Comparison of weight-for-height indices as a measure of adiposity and cardiovascular risk from childhood to young adulthood: the Bogalusa heart study. J Clin Epidemiol 2001;54:817 822. 35. Hall DM, Cole TJ. What use is the BMI? Arch Dis Child 2006;91:283 286. 36. Wang J, Thornton JC, Bari S et al. Comparisons of waist circumferences measured at 4 sites. Am J Clin Nutr 2003;77:379 384. 37. Ross R, Berentzen T, Bradshaw AJ et al. Does the relationship between waist circumference, morbidity and mortality depend on measurement protocol for waist circumference? Obes Rev 2008;9:312 325. 1068 VOLUME 19 NUMBER 5 May 2011 www.obesityjournal.org