Article Text

Download PDFPDF

Is mid-upper arm circumference in Dutch children useful in identifying obesity?
  1. Henk Talma1,
  2. Paula van Dommelen2,
  3. Joachim J Schweizer3,
  4. Boudewijn Bakker4,
  5. Joana E Kist-van Holthe1,
  6. J Mai M Chinapaw1,
  7. Remy A Hirasing1
  1. 1 Department Public and Occupational Health, Amsterdam Public Health, Child Health and Care Research, VU University Medical Centre, Amsterdam, The Netherlands
  2. 2 Statistics, Netherlands Institute of Applied Sciences TNO, Leiden, The Netherlands
  3. 3 Department of Paediatric Gastroenterology, Willem-Alexander Children’s Hospital/Leiden University Medical Centre, Leiden, The Netherlands
  4. 4 Department of Pediatrics, Reinier de Graaf Gasthuis, Delft, The Netherlands
  1. Correspondence to MD Henk Talma, Department Public and Occupational health, Child Health & Care Research, VU University Medical Centre, Amsterdam Public Health, Amsterdam, 1007 MB, The Netherlands; h.talma{at}vumc.nl

Abstract

Background Mid-upper arm circumference (MUAC) is suggested as being a valid measure in detecting overweight/obesity in children and adolescents, due to the strong relation with weight. We examined this relation and compared MUAC to body mass index (BMI) according to the International Obesity Task Force (IOTF) in children.

Methods Anthropometric data including MUAC were collected in 2009 by trained healthcare professionals in the context of the fifth Dutch Nationwide Growth Study, in a sample of 6167 children (2891 boys and 3276 girls) aged 2–18 years of Dutch origin. We propose MUAC SDS cut-off values for overweight and obesity, and compared MUAC with BMI IOTF in sex-specific and age-specific categories (2–5, 6–11, 12–18 years).

Results The area under the curve is used as a measure of diagnostic accuracy; the explained variance (R²) is good to excellent (0.88–0.94). Sensitivity ranges from 51.8% to 95.3% and specificity from 71.4% to 93.8%. Across age and gender groups, 65.1% to 89.0% participants are classified by both MUAC and BMI as normal weight, overweight or obese. We constructed three equations to predict weight using MUAC, with small differences between observed and predicted weight with an explained variance ranging from 0.88 to 0.94.

Conclusions Compared with BMI, MUAC is a valid measure for detecting overweight and obesity and thus a good alternative for BMI. When weight has to be estimated, it can be accurately predicted using MUAC. Based on our observations, we recommend developing diagrams with international (IOTF) cut-offs for MUAC SDS similar to BMI.

  • comm child health
  • epidemiology
  • health services research
  • obesity
  • adolescent health

Statistics from Altmetric.com

What is known about this topic?

  • There is little evidence about the relation between mid-upper arm circumference (MUAC) and overweight/obesity in children/adolescents, although it is suggested to be a good alternative for body mass index (BMI).

  • A strong relationship between MUAC and weight is suggested.

What this study adds?

  • MUAC is a good alternative for BMI in determining overweight/obesity in children/adolescents, when weight and/or height cannot be accurately measured.

  • The linear relation between weight and MUAC offers the possibility to predict weight.

Introduction

Mid-upper arm circumference (MUAC) is a valuable tool for screening nutritional state in children.1 2 Of all anthropometric measures, MUAC is the easiest to obtain, as well as being simple, practical and reliable.3–5 Measuring tapes are affordable and portable. MUAC is typically used in developing countries to diagnose and monitor underweight in children aged 6–60 months with a single cut-off value (12.5 cm; −2 SDS; WHO)6 because it is difficult to obtain reliable weight-for-length (2 years) or weight-for-height (>2 years) (WtfL/WtfHt), and the exact age. Special cut-offs, predicting risk of death, are suggested for infants (<12 months) with severe acute malnutrition.3 7 MUAC, and MUAC-for-age, provides information comparable to WtfHt in surveillance of populations.1 4 In infants, MUAC is even more sensitive and outperforms WtfHt in detecting malnutrition.8 9 The use of MUAC-for-age–SDS curves is recommended, preferably based on international data.4 10 Additionally, ethnic differences in MUAC cannot be ignored.10 11 Most studies on MUAC have been performed in undernourished and sick children at increased risk for mortality making comparison difficult. Studies in healthy children are scarce.10 MUAC is also considered a valuable tool in detecting overweight/obesity in children,3 12 although there are very few, nationally representative studies.13–18 MUAC cut-off values for overweight and obesity have been developed for children aged 1–3 and 4–6 years.3 12 Previous studies, in children over 5 years of age,3 13 14 19 20 showed that MUAC is a useful indicator for obesity screening due to the strong correlation with weight (Wt).4 10 13 14 19 20 MUAC is suggested as a measure in epidemiological studies as well as in clinical practice. Similar to other researchers,10 we searched for MUAC reference data from before the obesity epidemic. The only available study was the Dutch rural municipality study on MUAC—the Oosterwolde study—which is not nationally representative.21–23 This growth study in 2351 healthy Dutch children was carried out in 1979–1980.

Preventive Child Public Health (CPH) in the Netherlands is a unique system, incorporating Well Baby Clinics and municipal health services covering over 90% of children aged 0–18 years. One of the goals of CPH is the early detection of overweight/obesity.24 Normative sex-specific WtfHt growth diagrams are used25 and body mass index (BMI) is calculated according to the International Obesity Task Force (IOTF).26 27 All parents and/or children are advised about a healthy lifestyle, with overweight children being counselled by CPH and obese children being referred to a paediatrician. Because of the moderate accuracy of BMI, the individual’s physiognomy is crucial in establishing obesity, and this affects both the counselling of overweight children and the referral to a paediatrician of obese children.24

In 2008, it was decided to measure MUAC in the fifth Dutch Nationwide Growth Study (0–21 years; 2009) to investigate the potential of MUAC to determine overweight/obesity. Especially in epidemiological studies, we experienced increasing difficulty in persuading participants to undress in order to obtain accurate weight. We theorised that due to the strong correlation between MUAC and weight,10 28 and the observation of a possible linear relationship,29 that MUAC could also be a good alternative for estimating weight in clinical practice where children cannot be reliably weighed, for example, prescribing medication in an acute medical situation.

The aim of this study is to provide national representative data of MUAC in Caucasian children. We propose sex-specific MUAC SDS cut-off values for overweight and obesity in children aged 2 to 18 years, using the Oosterwolde study data as reference.21 22 The second aim is to compare MUAC with BMI in a contemporary national sample to investigate the potential of MUAC in assessing overweight and obesity.

A third aim is to provide equations for estimating weight (kg) using MUAC.

Methods

General

In the Netherlands, a national growth study is carried out once every 10 to 15 years (previous studies in 1955, 1965, 1980, 1997). These growth studies are cross-sectional and performed in a representative sample of Dutch children. Details on the design of the Dutch Growth Studies have been published previously.25 For the first time in 2009, MUAC was measured in addition to the usual anthropometric data. The Medical Ethical Review Board of Leiden University Medical Centre approved the study protocol and decided that written informed consent was not needed. MUAC data were collected in 6167 Caucasian children (2891 boys and 3276 girls) aged 2–18 years of Dutch origin. The children 2–3 years old were usually measured by CPH. The older participants were specifically invited for this study by CPH and measured at schools and at a festival.25 Data were collected and analysed anonymously.

Measurements

Trained healthcare professionals performed the measurements in the Dutch growth studies.25 Standing height was measured barefooted to the nearest 0.1 cm. Children were weighed (wearing underwear only) on calibrated mechanical or electronic step scales, with weight being rounded off to the nearest 0.1 kg.25 BMI was calculated as weight/height2 and classified according to the extended IOTF cut-off points, corresponding to a BMI at 18 years, for normal weight (BMI <25 kg/m²), overweight (BMI 25 to <30 kg/m²) and obesity (BMI 30 kg/m²).26 27 MUAC was measured midway the acromion and the olecranon in the non-dominant arm to the nearest 0.1 cm. Healthcare professionals were specifically trained and used identical measuring tape. MUAC SDS was calculated using the Oosterwolde study data.5 21 22 Age was determined to the nearest 0.01 year. Children with diagnosed growth disorders and those on medication known to interfere with growth were excluded.25

Statistical analyses

As only the percentiles of the Oosterwolde Study data were provided, LMS values (skewness curve (L), median curve (M) and coefficient of variation curve (S)) were needed to calculate MUAC SDS.30 The Newton-Raphson method was applied to convert the percentiles (P3, P10, P25, P50, P75, P90, P97) by age (per half year) and sex to LMS references, which enables each measurement to be converted into SDS. The difference between the percentiles based on the SDS and the original percentiles was at most 0.09 cm in boys and 0.10 cm in girls.

(SDS of measurement x is calculated as ((x/M)^L−1)/LS (when L ≠ 0) or ln(x/M)/S (when L=0). This SDS expresses the measurement in relation to the population in units of SDs above or below the median.)

Receiver operating characteristic (ROC) analysis was used to test the agreement between MUAC and BMI IOTF in classifying children/adolescents as normal weight (including underweight), overweight (excluding obesity) or obese. For the purpose of the analysis, participants were divided into sex-specific and age-specific child (2–5 and 6–11 years) and adolescent (12–18 years) groups.

As a measure of diagnostic accuracy, we used the area under the ROC curve. The ROC curve plots the sensitivity, that is, percentage overweight (including obesity) children with elevated MUAC SDS score (>1.3),21 22 against the 1−specificity, that is, percentage healthy weight children with a normal MUAC SDS score (1.3). The explained variance is given by R². The AUC is a way to reduce ROC performance to a single value representing expected performance, varying between 0.5 and 1.0, with 1.0 indicating a perfect test and ≤0.5 a worthless test, with higher values indicating a better predictive model. The categories used to summarise accuracy in ROC analysis were as follows: 0.9–1, excellent; 0.8–0.9, good; 0.7–0.8, fair; 0.6–0.7, poor; 0.5–0.6, fail.20 Linear regression analysis was used to determine predicted weight by using MUAC, and to determine significant differences in mean MUAC between predicted and observed weight. First, we applied linear regression analyses with weight as dependent variable and MUAC as independent variable. Second, we added sex to the model. Third, we added age and (age–mean age)² to the model. Analyses were performed in SPSS V.20 (linear regression analysis, analysis of variance, descriptive statistics).

Results

Table 1 shows the LMS values of the MUAC according to the Oosterwolde study data21 22 and the +1.3 and +2.3 SD cut-off points. These cut-off values for MUAC were chosen in agreement with approximately the same percentage overweight/obesity by BMI; 12.8% of Dutch boys and 14.8% of Dutch girls aged 2–21 years were overweight including obesity, and 1.8% of the boys and 2.2% of the girls were classified as obese.25

Table 1

LMS values (skewness curve (L), median curve (M) and coefficient of variation curve (S)) of mid-upper arm circumference according to the Oosterwolde study21 22 and the 1.3 and 2.3 SD cut-off values in boys and girls

Figure 1 shows the ROC and AUC (area under the ROC curve) curves based on cut-off points of the predicted probabilities from a logistic regression model according to IOTF overweight (including obesity) as an outcome measure and the MUAC SDS according to the Oosterwolde study 22 as independent variable.

Figure 1

Receiver operating characteristic (ROC) curves per gender and age categories. AUC, area under the ROC curve.

The AUC of the MUAC SDS to predict participants’ weight status ranged from 0.88 to 0.94 (high).20 The AUC was ‘excellent’ for boys and girls in all age groups (boys 2–5 years 0.90, both boys and girls 6–11 years 0.94, boys 12–18 years 0.94 and girls 0.93), and ‘good’ (0.88) for girls aged 2–5 years.

Table 2 shows sensitivity and specificity of MUAC >1.3 SDS21 25 in predicting overweight (including obesity). Sensitivity ranged from 51.8% to 95.3%, and specificity from 71.4% to 93.8%. Almost all sensitivities and specificities are high (>70%), except for sensitivity between 2 and 5 years of age. Therefore, half of the children in this age group with overweight or obesity according to BMI were not detected by MUAC.

Table 2

Sensitivity and specificity of mid-upper arm circumference >1.3 SDS in predicting overweight (including obesity) classified by Body Mass Index International Obesity Task Force26

Table 3 presents the number of participants per age category and gender. The percentages of individuals classified by MUAC SDS as well as by BMI IOTF in the same weight category range from 65.1 to 89.0 across age and gender groups (65.1% to 89.0% in boys, and 73.4% to 87.4% in girls; detection rate 78.8%). The percentages of participants with a normal BMI and an overweight MUAC (up to 16.4% in boys and 12.3% in girls), or an obese MUAC (up to 9.3% in males and 4.4% in females) were significant. A normal MUAC SDS in combination with an overweight/obese BMI was rare.

Table 3

Agreement between cut-offs for mid-upper arm circumference (MUAC) SDS21 and Body Mass Index International Obesity Task Force26 and number of participants per age category

Table 4 presents the regression coefficients to estimate weight (Wt) from only MUAC (equation 1), MUAC and sex (equation 2), and MUAC, sex and age (equation 3). The explained variance (R²) was good to excellent (0.88–0.94).20 The regression coefficients can be used to predict weight using MUAC, sex and age. The difference in Akaike Information Criterion between the first and second, and between the second and third regression models were respectively 154 and −4365.

Table 4

Results of stepwise linear regression analyses with weight (kg) as dependent variable, and mid-upper arm circumference (MUAC) (step 1), sex (step 2) and age (step 3) as independent variables in Dutch children aged 2–18 years and 95% CIs (n=6167)

Based on the regression coefficients (B) and intercept (table 4), we derived the following equations:

  1. Wt = 4.429 × MUAC − 57.867

  2. Wt = 4.435 × MUAC − 2.155 × Sex − 56.837 (females = 1; males = 0)

  3. Wt = 2.720 × MUAC − 1.925 × Sex + 1.765 × Age + 0.045 × Age² − 38.403

Age² = (Age − meanAge)² − Age = from 2 to 18 years − meanAge = 10.0 years

The mean and SD of the three regression models were respectively 0 (6.8), 0 (6.7) and 0 (4.8).

For example, a 6-year-old girl with an MUAC of 19.7 cm (sex=1, age=6, age centered2=(6−10)2=16) has a predicted weight of 2.720 × 19.7 − 1.925 × 1 + 1.765 × 6 + 0.045 × 16 – 38.403 = 24.6 kg.

Table 5 shows the differences in mean observed weight and predicted weight by using MUAC based on equation 3. In our example, the difference in mean observed weight and predicted weight is 0.667 kg (table 5).

Table 5

Differences in mean observed weight and predicted weight by age groups and gender

Discussion

We conclude that, compared with BMI, MUAC is a valid measure to determine overweight/obesity and appears to be a good alternative for BMI in children/adolescents 6 years onwards.3–5 MUAC can be especially useful when weight and/or height are difficult to obtain. To the best of our knowledge, this is the first national representative study on MUAC in Europe,25 with previous studies having mostly been performed in developing countries, the USA (NHANES)15–18 and Brazil. In the Netherlands, only one previous MUAC study has been performed in a healthy population (0–18 years) in a rural municipality in 1979/198021 22 and three very small-scale studies in preschoolers.12

The presented ROC–AUC curves show a good to excellent agreement between weight status categories based on BMI IOTF and our proposed MUAC SDS categories (1.3, >1.3 to <2.3, >2.3) in children and adolescents. Our results show a better agreement between BMI and MUAC than in a Turkish study in children (AUC ranged from 0.65 to 0.94).14

We constructed three equations (MUAC, MUAC and sex, and MUAC, sex and age) to predict weight based on MUAC with small differences in the mean in predicted and observed weight.21 29 These equations, especially the third, are useful in situations where weight has to be estimated, for instance when children are confined to bed and medicine has to be prescribed.

Child Public Health and obesity

In determining weight status in Dutch Child Public Health, the 1980 WtfHt and the BMI IOTF diagrams are used as normative.26 27 Overweight and obesity is eventually determined by experienced physicians by clinical judgement of physiognomy or body shape, because of poor correlation of BMI with percentage body fat (%BF),31 32 its moderate level of accuracy32 and low sensitivity.24 33–35 In our study, sensitivity of MUAC compared with BMI IOTF is low in children 2–5 years old, resulting in a considerable number of false positives to be checked by clinical judgement, to avoid overtreatment and stigmatisation. Obesity is associated with diabetes mellitus type 2, metabolic disturbances and inflammation, and is a serious public health issue in children/adolescents.36 37 Note that accurate assessment of weight status is crucial in advising and counselling by CPH. When BMI and MUAC agree on weight status (normal weight, overweight, obesity), this could reinforce the predictive value in selecting children with the highest health risks.38 The clinical judgement of obesity could become some less crucial. The major concern in CPH are the false negatives by BMI, the children with an increased %BF but normal weight (BMI <25 kg/m²) and with increased health risks.34 Of interest are also the overweight and obese children by BMI and a normal weight MUAC to determine to which extent they are the false positives by BMI or the false negatives by MUAC. The same applies for the overweight and obese children by MUAC and a normal weight BMI. Further research is necessary to clarify this, compared with %BF.

Performance of MUAC

With a detection rate of 78.8% of overweight/obese children, the number of false positives in our study is in line with previous studies. As current treatment of obesity is promoting a healthy lifestyle and this is important for all children, we believe the 9.3% false positives is tolerable (Table 4).13 Our study showed a strong correlation between MUAC and BMI, as in previous studies.10 14 The classification accuracy of MUAC was higher for BMI than for %BF in a study among 978 healthy children.20 Due to the lack of relevant studies on MUAC, it remains unclear whether, compared with %BF, BMI or MUAC is more accurate.13 14 20 The possibility of ethnic differences also needs further study. Similar to the WHO MUAC SDS diagrams in pre-schoolers,6 MUAC SDS diagrams are urgently needed in children over 5 years, including cut-offs for overweight and obesity, similar to BMI IOTF. There are very few MUAC studies in healthy children/adolescents available, with most data on MUAC being collected in the US NHANES studies.15–18 We were not aware of MUAC data from before 1988 to be able to establish a possible secular trend. Moreover, combining BMI and MUAC could improve the detection of overweight/obesity in children. This needs further research. MUAC could be the measurement of choice in determining overweight/obesity when weight and/or height cannot accurately be obtained. The expected secular trend in MUAC during the obesity epidemic needs confirmation, as well as the correlation between MUAC and FM, and MUAC and FFM.10 Although there are no established MUAC cut-offs for overweight and/or obesity in adults (18 years and older), we nevertheless propose the construction of international MUAC SDS diagrams for children, thereby enabling international comparability. BMI is essential in diagnosing obesity, but additional anthropometric indices are needed to describe body fat distribution more accurately.14

Strengths and limitations

The main strength of this study is using the same measurement protocol in all Dutch growth studies, and its representativeness in a large population of Dutch/Caucasian origin with sufficient statistical power, and the comparability with international studies by using BMI IOTF. We trained the healthcare professionals in measuring MUAC, and all health professionals used identical measuring tape.5 Due to the cross-sectional design, it is not possible to establish causal relationships. The main limitation of this study is the lack of a gold standard for percentage body fat to compare with, besides BMI.

Conclusion and recommendations

Compared with BMI, in case height and/or weight are not available, MUAC is a good alternative in detecting overweight/obesity in children over 6 years. Future studies should assess the validity of MUAC compared with gold standards for assessing %BF. We further suggest research on the predictive value of an increased MUAC for health risk, possible ethnic differences in MUAC cut-off values and the validity of the presented MUAC equations in predicting weight. Finally, we recommend developing diagrams with international cut-offs for MUAC-for-age standard deviation score  similar to BMI.

Acknowledgments

We thank all the children, their parents and the Child Public Health workers who participated in this study. We also thank Mariëtte Hoogsteder PhD for being a temporary supervisor.

References

Footnotes

  • Contributors HT wrote the manuscript and performed basic statistics. PvD analysed the data, wrote the statistical paragraph, critical reviewer. JJS conceived part of the study, critical reviewer. BB: critical reviewer. JEK-vH: supervisor, critical reviewer. JMMC: manager CH and CR, final approval, critical reviewer. RAH: conceived the study, critical reviewer.

  • Funding The Dutch Ministry of Health, Welfare and Sport (grant nos. 310434, 312617 and 315319).

  • Competing interests None declared.

  • Patient consent Not required.

  • Ethics approval The study was approved by the Medical Ethical Review Board of Leiden University Medical Centre.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.