The Journal of Health, Population and Nutrition icddr,b ISSN: 1606-0997 EISSN: 2072-1315 Vol. 28, Num. 3, 2010, pp. 221-229

Journal of Health Population and Nutrition, Vol. 28, No. 3, May-June, 2010, pp. 221-229

Article

Prediction equations for body-fat percentage in Indian infants and young children using skinfold thickness and mid-arm circumference

Bandana Sen¹ , Kaushik Bose² , Saijuddin Shaikh¹ , and Dilip Mahalanabis¹

¹ Society for Applied Studies, CF-198, Salt Lake City, Sector I, Kolkata 700 064, India and
² Department of Anthropology, Vidyasagar University, Paschim Midnapore 721 102, India

Correspondence Address: Dr. Dilip Mahalanabis, Society for Applied Studies, CF-198, Salt Lake City, Sector I, Kolkata - 700 064, India

Code Number: hn10029

Abstract

Keywords: Anthropometry; Body mass index; D ₂ O dilution; Fat-mass; Infants; Mid-arm circumference; India

Introduction

Studies in developed countries have shown that slow growth during the first two years of life is associated with an increased risk of coronary heart disease in adults, independently of birthweight ^{[1],[2],[3],[4]} . However, postnatal rapid gain in weight has also been suggested to be a risk factor for later obesity, elevated blood pressure in adolescent males, impaired glucose tolerance in young adults, and increased mortality from coronary heart disease ^{[5],[6],[7],[8],[9],[10],[11]} . A recent analysis of longitudinal data in low-income and middle-income countries has shown that stunting in the first two years of life leads to shorter height in adults, lower attained schooling, and decreased birthweight of offspring ^[12] . The analysis has also shown that stunting between 12 and 36 months of age predicts poor cognitive performance and/or lower school grades attained in middle childhood ^[12] . Further, published work from high-income countries suggests a consistent association between intrauterine growth restriction and an increased risk of several chronic diseases in adulthood, such as coronary heart disease, diabetes, and hypertension ^{[13],[14],[15],[16]} .

In light of these findings, precise measurement of body composition, such as fat-free mass (FFM) and fat-mass (FM) in infancy and early childhood, assumes importance. Many methods for measuring body-fat percentage are sophisticated and only suitable for the research setting ^{[17],[18],[19],[20],[21],[22]} . Among the large number of body-composition assessment methods in common use, anthropometry, includ-ing skinfold measurements, is simple, portable, and cost-efficient field methods.

We need a technique that provides reliable and valid estimates of body-fat in infants and young children that are non-invasive and are suitable for use in the community and outside fixed facilities. Sufficiently accurate measurement of length of in-fants and young children for use in equations for body composition is difficult to carry out in the field, largely because portable measuring instruments are inaccurate and one needs the child's cooperation. Similarly, for measuring weight, one needs to carry weighing scales to the field. On the other hand, with adequate training, health workers can measure midarm circumference and skin-fold thickness with portable equipment and with less burden on participants. In the present study, we developed two new equations for estimating the body-fat percentage (one for girls and the other for boys) on 6-24-month old infants using skinfold thickness (biceps, triceps, and suprailiac), mid-arm circumference, and age. We used the D ₂ O dilution technique as the reference method. We also validated these equations on an independent sample of infants to estimate the body-fat percentage and compared them with the estimates derived by the reference method.

Materials and Methods

Subjects

The study was conducted among the urban poor children attending a well baby clinic of a large charitable government hospital in the city of Kolkata, India. Sixty-nine children, aged 6-24-months, participated in the study. The socioeconomic and demographic features of the families are presented in [Table - 1]. The equations were developed on 46 (27 girls and 19 boys) consecutively-enrolled infants and young children, and then these predictive equations were validated on 23 infants (11 girls and 12 boys) in the same age-group and from the same community enrolled for this purpose. It may be noted that regression equations fitted on a sample can only be validated on an independent sample. The sample-size was largely determined by the funds made available by the Society for Applied Studies Trust for this research project. Written informed consent was obtained from their parents, and anonymity of the subjects was strictly preserved. The prediction groups of infants were initially recruited, and the prediction equations were developed. The validation group was recruited after that. The plan was to recruit a larger number for validation. High expense for the D ₂ O test restricted the numbers.

Measurements

We used the D ₂ O dilution technique for measuring total body-water (TBW). We measured length, weight, skinfold thicknesses (biceps, triceps, and suprailiac), and mid-arm circumference of these in-fants ^{[23],[24],[25],[26],[27]} .

Length: Recumbent length was measured with a wooden measuring-board as the age was less than two years. Two examiners were required to correctly position the infant and ensure the accurate and reliable measurement of length. The subject was placed with the face upward and the head towards the fixed end and the body parallel to the long axis of the board. The shoulder-blades were rested against the surface of the board. One examiner applied gentle traction, brought the crown of the child's head into contact with the fixed headboard, and positioned the head so that the Frankfort plane was vertical. The second examiner held the subject's feet, without shoes, toes pointing directly upward and keeping the subject's knees straight, and brought the movable footboard to rest firmly against the heels. The reading was taken to the nearest 0.1 cm. Only the left leg was positioned for the measurement when the subject was restless ^[9] .

Skin-fold thickness

Measurements of skinfold thickness were performed in duplicate on the left side of the body on all subjects with a Lange Skinfold Calliper (Cambridge Scientific Instruments, Cambridge, MD). All measurements were undertaken according to the Anthropometric Standardization Reference Manual ^[26] . The methods have also been described and illustrated by Cameron ^[25] .

Triceps: Vertical fold was measured at the posterior midline of the upper arm, halfway between the tip of the shoulder and tip of the elbow, the elbow remaining in an extended and relaxed position.

Biceps: It was measured as the thickness of a vertical fold on the front of the upper left arm and directly above the centre of the cubital fossa, at the same level as the triceps skinfold.

Suprailiac: It was measured in the midaxillary line immediately superior to the iliac crest. The skinfold is picked up obliquely just posterior to the midaxillary line and parallel to the cleavage lines of the skin ^[26] . The suprailiac is not predominantly 'vertical' but rather follows the natural fold of the skin.

Mid-arm circumference

A plastic non-elastic tape was used for measuring mid-arm circumference. The measurement was taken at the midpoint of the upper left arm, be tween the acromion process and the tip of the olecranon. After locating the midpoint, the left arm was extended so that it was placed loosely by the side, with the palm facing inwards. The tape was wrapped gently but firmly around the arm at the midpoint. Measurements were taken to the nearest millimetre ^[24] .

Weight: Weight was measured nude, using an electronic platform balance (Avery India Limited, model no. L111A) with a precision of 10 g. The balance was checked regularly for accuracy using standard weights.

Age: Age was calculated as the difference between the date of study and the date of birth of the child (as shown on their birth certificates).

D ₂ O dilution method: Deuterium oxide (D ₂ O), water composed of a stable isotope of hydrogen, was used. The test was performed between 9 am and 2 pm in the study centre. Baseline saliva of children was collected using a disposable syringe without needle. The subject then drank D ₂ O (99.9%, Sigma) in a dose of approximately 50 mg/ kg of bodyweight ^[23],[28] . We weighed the D ₂ O dose in a disposable container using a precision balance, and the weight with an accuracy of 0.1 mg was recorded and used in calculating TBW. It was then mixed with 20 mL of distilled water, and the child drank it from a disposable syringe without needle. The empty container was rinsed with two consecutive lots of 15 mL of distilled water, and the child drank both the lots from the same disposable syringe. Time was allowed to equilibrate the D ₂ O into the body-fluid. No food and water was permitted during the equilibration period of three hours. At the end of the equilibration period, the follow-up saliva sample was collected and immediately stored at -20 ⁰C until transported to the St John's Hospital, Bangalore, India, for further analysis. The samples were analyzed for deuterium using a duel-inlet mass spectrometer (Europa Scientific, Crewe, UK) using the zinc reduction technique ^[27] . The results of D ₂ O concentration in the saliva samples were used for calculating the TBW using the following equation ^[29] :

TBW in moles (N2)=(F1ΧN1)÷(F2Χ1.041), where

F ₁ =dose in atom percentΧ10 ⁶ /100-150 ppm

F ₂ =ppm after dose-ppm before dose

N ₁ =Dose in g/molecular weight of D ₂ O (20.0274)

TBW in kg=(TBW in molesΧmolecular weight of H ₂ O (18.0153))/1000

The FFM was obtained as TBW in kg divided by an age and sex-specific hydration factor for FFM, and for this age-group, it ranged from 80.7 to 77.0 for boys and 80.7 to 78.0 for girls ^[30] .

Body-fat percentage={(Weight-FFM)/weight}Χ100

Regression equations

We used the fat percentage derived from the equation based on the D ₂ O dilution method as stated above for developing regression equations using a forward stepwise regression model. We fitted linear regression models with the three skinfold measurements (biceps, triceps ² , and suprailiac), mid-arm circumference, and squared values of age as predictor variables. The best fitted models with biceps and suprailiac as predictors, i.e. Model 2 for boys and Model 5 (with mid-arm circumference, age ² , triceps ² , and biceps as the predictor variables), for girls were chosen. For comparing the models, we used adjusted R ² values ^[31] . We examined the degree of agreement between the body-fat percentage derived by the new equations with those derived from the D ₂ O dilution equation on the infants and young children on whom the equations were developed. For developing the equations, we have added one variable at a time as predictor variables as stated earlier and examined the values of adjust ed R ² .

Statistical analysis

The Epi Info ^TM software (version 6) ^[32] and the Stata software (version 7.0) ^[33] were used. For anthropometric data, a software package based on the National Center for Health Statistics (NCHS) (Centers for Disease Control and Prevention-CDC) database as provided with the Epi Info software was used. These anthropometric calculations were based on the growth reference curves developed by the NCHS and CDC ^[34] . The best fit equations derived by regression models were first examined on the infants on whom the equations were developed. We then used these on a new set of infants in the same age-group, who also had the D ₂ O dilution test and compared the body-fat percentage derived by the developed equations and by the D ₂ O reference method. The percentage of fat derived by the new equations was plotted against those based on the equation for D ₂ O dilution by scattergrams, and the agreement was visually examined to note if the points lie close to the 'line of equality', i.e. the line at a 45 ⁰ angle. Furthermore, the differences in the body-fat percentage for each child of the validation set between those derived by the new equations and by D ₂ O dilution were evaluated for their deviation from zero by t-test. For reasonably symmetric distri-bution, we expect the range, mean±2 standard de-viation, to include about 95% of the observations, and we used this to indicate 95% limits of agree-ment ^[31] . To evaluate how well the methods are likely to agree for comparing groups, we used 95% confidence interval of the mean of the differences in each subject to examine the agreement between the two groups. For a good agreement, we expect that the range, mean of the difference ±2 standard error to indicate 95% limits of agreement within which group means will lie 95% of the time.

Ethical issues

Ethical clearance was obtained from the Ethical Re-view Committee of the Society for Applied Studies, Kolkata. The objective of the study was explained, and the written informed consent was obtained from their parents.

Results

The study children came from a relatively-poor community in a metropolitan city in India. The average family income was US$ 44 per month, and 65% of the families lived in one room. Fifty-six percent of the mothers had school education of six years or more. These features are comparable with the norms for the urban population in India. The mean (SD) of age, weight, length, body mass index (BMI) percentiles for age and sex, biceps, triceps, and suprailiac skinfold thickness, mid-arm circumference and z-scores for weight-for-age, length-for-age, and weight-for-length of the infants on whom the equations were developed (i.e. prediction group: 27 girls and 19 boys) and those in whom these equations were validated (i.e. validation group: 11 girls and 12 boys) are shown in [Table - 1]. Seventeen children (9 girls and 8 boys) in the prediction group and two children in the vali-dation group were aged less than 12 months. Oth-ers were aged 12-24 months. [Table - 2] shows the ad-justed R ² values for each model with the increasing number of predictor variables. Many other mod-els were also examined but not shown here. We note that the last model (i.e. Model 5) was the best fit for girls. For boys, Model 2 was the best fit model. Model 5 included biceps, mid-arm circum-ference, age ² , and triceps ² , and Model 2 included biceps and suprailiac as predictor variables. The regression coefficients (SE) of the chosen models based on fat-mass percentage derived by D ₂ O dilu-tion as the reference method are shown in [Table - 3]. We have taken these values to construct the final equations, which are:

Girls: body-fat percentage=-69.26+5.76ΧB -0.33ΧT ² + 5.40ΧM+0.01ΧA ² (1)

Boys: body-fat percentage=-8.75+3.73ΧB+2.57ΧS (2), where

B=biceps skinfold thickness in mm, T=triceps skin-fold thickness in mm,

S=suprailiac skinfold thickness in mm, M=mid-arm circumference in cm, and A=age in months.

The mean (SD) of the body-fat percentage calculated by the new equations was 17.11 (6.25) and 16.93 (6.02) for girls and boys respectively. Likewise, the mean (SD) of the body-fat percentage calculated by the D ₂ O dilution methods was 17.11 (7.25) and 16.93 (6.62) for girls and boys respectively. The 95% confidence limits of the mean of the difference in the body-fat percentage derived by D ₂ O and by the new equations were -1.45 to +1.45% for girls and -1.33 to +1.33% for boys. As expected, the mean of the difference in the body-fat percentage was zero.

Applicability of predictive equations

Using the new prediction equations (1 and 2), we calculated the fat percentage in an independent sample of 23 infants (11 girls and 12 boys) in the same age-group and from the same community. We also calculated the body-fat percentage on them using the D ₂ O dilution equation. The mean (SD) of the predicted body-fat percentage derived by the new equations was 20.15 (7.06) and 14.21 (4.27) and by the D ₂ O dilution was 19.22 (9.54) and 15.36 (4.49) for girls and boys respectively. The mean (SE) of the difference in the body-fat percent-age was -0.93 (1.98) for girls and 1.14 (0.70) for boys. The mean (SD) for the difference in the body-fat percentage was -0.93% (6.56) for girls and 1.14% (2.43) for boys. Distribution of data points along the line of identity (45 ⁰ ) between the body-fat percentage derived from the D ₂ O dilution method and by the new equation applied on the validated group is shown in [Figure - 1]. The Bland-Altman plots of the paired differences between the body-fat percentage derived by the D ₂ O reference method and the new equations for the validation group of girls and boys are plotted against average body-fat percentage derived by the D ₂ O method and respective new equations [Figure - 2] ^[35] .

Discussion

The purpose of this study was to develop predictive new equations for body-fat percentage using simple measurements, such as skinfold thickness and mid-arm circumference, for use in community nutritional studies of infants and young children in India. For infants and young children, a few regression equations to estimate the body-fat percentage derived from density are available ^[36] . The relationship between total body-density and skinfold thickness varies with age and sex ^{[37],[38],[39]} . Weststrate and Deurenberg, however, used age and sex-specific equations based on skinfold thickness to derive body-density and body-fat percentage ^[40] . In the infants we studied, the calculated body-fat percentage based on these equations (data not shown) was substantially lower than the ones de-rived by the D ₂ O dilution reference method and, as such, not suitable for this population.

We were able to validate these new equations on a small number of children and to show that the mean difference of the body-fat percentage was only-0.93 for girls and 1.14 for boys. In South Asian countries, there are no equations of this type for this age-group (6-24 months) for calculating the body-fat percentage. Our results showed that the proposed equations might be used for deriving the percentage of body-fat of infants to compare among groups. The number of children for validation was relatively small, which was largely limited by the cost of D ₂ O estimation. In the absence of any validated equations for infants in South Asia, these equations should, therefore, be useful until further validation is feasible with additional data. Due to relatively wide limits of agreement (i.e. relatively large standard deviation of the mean difference) with the reference method, these equations are less suited for the assessment of body-fat of individual infants. However, the 95% confidence limits of the mean difference of paired values in the validation set of the infants suggest that these equations should be useful for public-health and epidemiological studies.

The measurement of skinfold thicknesses at conventional sites may not capture some important components of body-fat, i.e. central fat and lower limb-fat, in infants. We speculate that adding measurements, such as waist-circumference, thigh- and calf-skinfolds and circumferences, may improve the limits of agreement for body-fat percentage with those derived by the reference methods. Sub-stantial changes in body-composition occur during infancy and puberty. Thus, assessment of body-composition in infants is more challenging than in adults. Barring cadeveric studies, measurements of body-composition are indirect and involve assumptions that may introduce error. Prediction equations based on anthropometric measures are derived from samples of healthy children and are dependent on several important assumptions, such as water and mineral composition of FFM, fat distribution, and patterns of skeletal maturation. Skin-fold thicknesses, particularly triceps, are also directly used as indicators of fatness. Reference data exist for individuals from one year of age through adulthood ^[40],[41] . The sum of triceps and subscapular skinfolds is also considered to be a good indicator of overall fatness, and reference data are available ^{[41],[42],[43]} . To use skinfold thickness directly to indicate fatness or leanness, we need reference data for infants which are not available.

To compare the body-fat percentage in the study infants with those of infants of similar age in developed countries, we plotted individual data-points of body-fat percentage derived by D ₂ O dilution in all the study infants against the mean reference data [Figure - 3] on healthy American infants of similar age derived by multi-component modelling ^[30] . We note that the body-fat percentage in both boys and girls are generally much lower than those found in healthy American infants of similar age-group. These apparently-healthy infants drawn from a relatively low socioeconomic stratum in a metro-politan city in India are generally very lean. This phenomenon largely reflects inadequate nutrition during infancy and early childhood and is likely to be a consequence of the well-known phenomenon of prolonged breastfeeding combined with inadequate weaning food of low energydensity.

Babies in South Asia have a lower birthweight than their Western counterparts, and studies have shown that body-fat is relatively preserved in these babies leading to their description as 'thin-fat' ^[44] . Further, adult populations in India have a higher percentage of body-fat for a given BMI than Western populations ^[45],[46] . Based on the above, it has been speculated that increased body-fat at birth tracks into adulthood depending on environmental constraints ^[47] . This study has generated quality data on fat and lean body-mass percentage in a group of infants from low- to middle-income families using a reference method based on the D ₂ O dilution technique. On examining the association of the percentage of fat in our study infants [Figure - 4] with leanness as indicated by their weight-for-length percentile, we note that 52 (83%) are below 25th percentile for weight-for-length by the NCHS standard who may be considered lean. Of these 52 infants, six (11.5%) had more than 25% body-fat. Therefore, nearly 12% of the lean infants had high fat percentage who may be considered to fall among the so-called thin-fat phenotype as described by Yajnik and colleagues ^[44] . In a critical evaluation of the phenomenon of thin-fat phenotype at birth, Muthayya and colleagues have shown that the thin-fat paradigm may not apply for babies small-for-gestational age ^[47] . Since the birthweights of the study infants are not known, approximately 30% of the babies born in this pop-ulation are known to have low birthweight ^[48] .

The equations we have developed for calculating fat percentage in infants and young children will be useful in studying growth and body-composi-tion in the first two years of life. Studying body-composition in this age-group has some relevance. There is good evidence that the trajectories of growth during infancy and childhood are a major risk factor for a group of chronic diseases that in-clude coronary heart disease and diabetes ^{[1],[2],[3],[4],[49]} . Although low birthweight is known to be associ-ated with increased rates of coronary heart disease and related disorders, such as stroke, hypertension, and type 2 diabetes in adulthood, recent findings indicate that impaired growth in infancy and rapid gain in the childhood weight are independently as-sociated with these chronic diseases of adults ^[1] . More specifically, low BMI at two years of age and high BMI at 11 years of age were shown to be as-sociated with coronary events. On the other hand, there is evidence to suggest that faster growth as indicated by percentile crossing, particularly in infancy is associated with metabolic syndrome in later life ^[50] .

Acknowledgements

The work was supported by a grant from the Society for Applied Studies Trust.

The authors thank Dr. Anura V. Kurpad for doing the D ₂ O estimation in the saliva samples at his laboratory at St. Johns Medical College, Bangalore. They also thank Md. Jakir Hossain for statistical as-sistance.

References

Copyright 2010 - Journal of Health Population and Nutrition

The following images related to this document are available:

Photo images