Statistical modelling of body composition in children and their parents

Franco Villoria, Maria (2009) Statistical modelling of body composition in children and their parents. MSc(R) thesis, University of Glasgow.

Abstract

Several methods of assessing body composition are presented and discussed in this thesis. The assessment of body composition has been a leading area of research for many years. The increasing prevalence of obesity in recent years has done nothing but stress the importance of having accurate and reliable methods which allow us to understand better the composition of the body itself, as well as identify any abnormal conditions of the same.
Data was provided by the Gateshead Millennium Study (GMS). This is a large cohort of over 1000 children, who were born in Gateshead, in the Northeast of England, between 1999 and 2000. The children have so far been followed up until age 7, when the last data collection was completed. Information about their parents (such as anthropometric measurements) was also recorded. This thesis analyzes the data of the children at age 7 as well as their parents.
A number of methods for determining body composition are currently in use. These vary from taking simple measurements of the body, such as weight or height, to more complicated methods for which sophisticated devices are needed. Various relationships and associations between the different characteristics of the human body have been established, drawn from results of numerous studies. Although these relationships may vary from person to person, as there are many other factors which might affect them (such as diet, genetics or lifestyle), they are fairly well established in adults. However, when assessing children, an important factor should be considered in addition, age. This is fundamental for the correct assessment of body composition in children. This thesis focusses on two more novel methods of determining body composition, bioelectrical impedance analysis (BIA) and bony frame, in addition to anthropometry (body mass index in particular) and skinfolds.
Parental body composition is thought to be influential on their children's body composition. Such influence does not necessarily have to be genetic, but might indirectly come from the eating habits and lifestyle of the family as a whole. Another factor which is believed to have an influential role is social class. Social differences might be reflected in body composition differences, with people from lower social classes being fatter. Both factors are explored in this thesis.
Chapter 1 gives a general introduction on how body composition can be assessed and describes the data set and aims of the project. These include processing anthropometric (height, weight, body mass index, skinfold thicknesses, waist circumference, bony frame) and impedance data for children and their parents, as well as comparing and validating different methods of assessing body composition.
Chapter 2 provides a review of some relevant published literature and methodology concerning assessment of body composition in children and adults. The principles on which bioelectrical impedance analysis is based are described and different approaches suggested in the past for analyzing bony frame data are reviewed.
Chapter 3 presents the anthropometric data collected on children and their mothers in the GMS. The LMS method was used to calculate standard deviation scores for children's anthropometric data. The results reflected an increase in height, weight, waist circumference, BMI and skinfolds (triceps and subscapular) with respect to reference data (UK 1990 children). In particular, classification criteria based on body mass index and skinfold thicknesses are discussed. Children were classified under the UK National and the IOTF body mass index criteria. The resulting classifications differed, especially for boys. Two different equations were applied to predict %fat from skinfolds, but discrepant results were obtained. The relationship between mothers and children's body mass index was explored. Although it was found to be significant, the correlation was weak.
Of particular interest for this project is bioelectrical impedance analysis (BIA), explored in Chapter 4. This approach is presented as a way of studying the fat and lean mass components of the body separately. A novel methodology for deriving lean and fat indices in children, adjusted for height and age, is discussed. These methods were derived on another data set and are applied here to the GMS children. The results are encouraging, although it is a matter of concern that the variance of the fat index was smaller than expected. Both indices are correlated with weight, waist circumference, BMI and skinfolds. Fat index yields considerably higher correlation with skinfolds than lean index. The relationship between fat index and other measures of fatness is stronger for girls, especially for BMI; while for boys the correlation coefficients with BMI are about the same for both indices, for girls the correlation with fat index is clearly higher. The methodology was then adapted to derive similar indices for the children's mothers. The resulting indices, especially the fat index, are highly correlated with the body mass index. Mothers and children's lean and fat indices are positively (but not strongly) correlated.
Chapter 5 explores bony frame data. Faced with the lack of a gold standard method, an alternative way of analyzing these data is proposed, taking the average of the internally standardized measurements. The results from a principal components analysis supported this idea. Average limb and trunk measurements were considered separately but this approach did not seem to be better than taking the overall average. Correlation coefficients between bony frame and height, weight and BMI are all high. Waist circumference and skinfolds, which are measures of fatness, are also correlated with bony frame. In general, correlation coefficients are higher for girls. For boys, the correlation of bony frame with lean index is higher than with fat index, while for girls the opposite occurs.
Finally, Chapter 6 summarizes the results and main findings, pointing out the limitations in the methodology used and suggesting areas of study which should be considered in the future.
Bioelectrical impedance analysis has been reported to be a valid and reliable method for assessing body composition in previous studies. The results presented in this thesis are very encouraging. In general, fat and lean indices seem to perform fairly well for both children and adults; the relationships between these indices and various anthropometric measurements agree in sign and magnitude with the expectations. However, there are limitations one should be aware of. The equations for calculating the lean and fat indices for children were derived on a specific age range (7-11 years) and therefore, might not be suitable for children aged outside that range. Also, when applying this method to the GMS children, the variance of the fat index was found to be smaller than expected. The reason for this is not clear yet, and further investigation should be done. In adults, the equations for calculating these indices were derived using solely women's data. The hydration and resistivity constants underpinning the method might differ for males, and also for different age ranges.
The results obtained from analyzing the bony frame data are unclear. The high correlation with fatness, which, in principle, was not expected, could mean two different things: either an actual relationship between bony frame and fat mass, or simply the fact that bony frame calculation is being confounded with fat, and therefore the results would not be reliable. Further work on this field should be done, so that these questions can be answered and an appropriate and reliable method for adjusting for bony frame can be developed.

• Item Type: Thesis (MSc(R))
• Qualification Level: Masters
• Keywords: body composition, bioelectrical impedance analysis, bony frame
• Subjects: Q Science > QA Mathematics
• Colleges/Schools: College of Science and Engineering > School of Mathematics and Statistics > Statistics
• Supervisor's Name: McColl, Prof. John
• Date of Award: 2009
• Depositing User: miss maria franco villoria
• Unique ID: glathesis:2009-589
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 13 Feb 2009
• Last Modified: 10 Dec 2012 13:20
• URI: https://theses.gla.ac.uk/id/eprint/589