Children and adolescents living with HIV undergoing highly active antiretroviral therapy (HAART) often exhibit changes in body fat such as lipoatrophy (loss), lipohypertrophy (central accumulation), or a combination of both.1 These changes are associated with metabolic abnormalities such as mitochondrial toxicity, dyslipidemia, and insulin resistance, which can increase the risk of cardiovascular and cerebrovascular diseases.2 Moreover, changes in body fat can affect body image perception, cause poor adherence to HAART, and compromise quality of life.3

The prevalence of lipodystrophy ranges from 26% to 55% among children and adolescents living with HIV.4–6 This variability can be attributed to the subjectivity of visual inspection – mainly used for diagnosis – and to differences in time of exposure and in type of HAART. Objective methods have been used for measuring body fat.7,8 Air-displacement plethysmography is a precise densitometric method for evaluating body fat, while dual energy X-ray absorptiometry (DXA) is an imaging-based alternative technique. DXA is a safe and accurate method,9 especially when a more comprehensive compartmental analysis of body composition is required. However, both methods are expensive and require sophisticated infrastructure and human resources, thus rendering the application of these assessments difficult under conditions of limited resources.

Anthropometry is a noninvasive, low-cost and easy-to-use method that could be applied as an alternative to estimate body fat. Skinfold thickness studies have been performed to estimate body fat in the general population because of the increase in obesity.10 However, it is unknown whether the commonly used equations of Lohman11 and Slaughter12 are valid for children and adolescents living with HIV. Studies have used these equations indiscriminately to predict body fat13 or skinfold to evaluate thickness of subcutaneous body fat.14,15 However, the predictive value of anthropometric measurements to estimate body fat using specific equations requires further investigation. The aim of this study was to assess the validity of traditional anthropometric equations and to develop predictive equations of total body fat and trunk fat for children and adolescents living with HIV based on anthropometric measurements.

The sample consisted of 48 children and adolescents living with HIV. Most of the participants had white skin color (56.3%), were from families with a household income of up to two minimum wages (47.9%), and were in the first three stages of sexual maturity (66.6%). Regarding clinical condition, most participants were asymptomatic or had mild symptoms (52.1%), had moderate immunosuppression (58.3%), and received non-protease inhibitor HAART (58.3%). Most adolescents were eutrophic (87.5%), but girls exhibited greater skinfolds and body fat by DXA than boys (Table 2).

As can be seen in Table 3, subscapular skinfold, the sum of four skinfolds and abdominal skinfold were linearly correlated with total body fat. For trunk fat, subscapular skinfold showed a higher correlation than BMI and abdominal skinfold.

Table 4 shows the prediction models of total body fat and trunk fat. Model 1 explained 85% of total body fat measured by DXA based on subscapular skinfold (standardized beta coefficient [βstandardized]=0.38), sum of four skinfolds (βstandardized=0.71), abdominal skinfold (βstandardized=−0.33), sex (βstandardized=−0.20), and height (βstandardized=0.26). The SEE of this model was low (1.42kg). Model 2 explained 80% of the variance in trunk fat, with an SEE of 0.49kg, based on subscapular skinfold (βstandardized=0.63), age (βstandardized=0.31), and sex (βstandardized=-0.21).

There was no difference between the total body fat values measured by DXA and those estimated with Model 1 (Table 5). Similarly, no difference was observed between trunk fat values measured by DXA and those estimated with Model 2 (p>0.05). In addition, Lin's concordance correlation coefficients were of moderate to substantial magnitude (Model 1=0.93; Model 2=0.90). However, there was a significant difference between the body fat values predicted with the equations of Lohman and Slaughter and the DXA values.

Fig. 1 shows linearity of measured body fat and body fat predicted with our models. Bland–Altman analysis of residuals is shown in Fig. 2. A random distribution of residuals was only found for total body fat (x2=1.93 and p=0.1649), but not for trunk fat (x2=15.51 and p<0.001). The same was observed for the equations of Lohman and Slaughter. Internal validation using 5000 bootstrap samples produced a corrected R2 of 85% and 80% for Models 1 and 2, respectively.

Fig. 1.

Association between body fat measured by DXA (total [kg] and trunk [kg]) and predicted by the equations developed (Total BFmodel1[a],Trunk BFmodel 2[b] and Total BFLohman[c] and Total BFSlaughter[d]) in children and adolescents living with HIV.

(0.3MB).

Fig. 2.

Analysis of residual scores of total body fat [BF] (a), trunk body fat (b) measured by DXA, prediction models and by Slaughter (c) and Lohman (d) equations in children and adolescents living with HIV.

(0.38MB).

The main findings of this study were that subscapular skinfold, the sum of four skinfolds, abdominal skinfold, sex, and height explained 85% of total body fat. Moreover, subscapular skinfold, age, and sex explained 80% of trunk fat in children and adolescents living with HIV. Although correlated with DXA-measured body fat, the traditional equations of Lohman and Slaughter underestimated total body fat by 8%. These findings demonstrate that our models based on anthropometric measurements can be used by clinicians and healthcare professionals as an objective, noninvasive and inexpensive alternative to assess body fat in children and adolescents living with HIV.

Models 1 and 2 developed in the present study can be used to estimate total body fat and trunk fat, respectively, showing high predictive power and agreement with the DXA method. In addition, a low error of estimation was found. These models represent a practical tool for the assessment of body fat in the clinical setting, although some variability in body fat may occur due to growth and sexual dimorphism, which interfere with the prediction and interpretation of the results.25,26 However, it is important to include these variables in the equations since they are associated with lipodystrophy.1,14

Studies on HIV conducted in developed countries have frequently used DXA to analyze body fat patterns and the association with cardiometabolic diseases.8,27 However, the high cost of the equipment and the need for adequate infrastructure and personnel restrict the use of this method for body fat assessment when resources are limited. The development of specific anthropometric equations for children and adolescents with HIV is important for monitoring body fat (Model 1) because of the changes that result from HAART and HIV infection itself.1 Although no cut-off points that define abnormalities are available, it is evident that the children and adolescents evaluated in this study are suffering from a reduction in body fat since negative median z-scores were found, especially in boys (Table 2), indicating a lipoatrophy phenotype.

Lipohypertrophy consists of the accumulation of adipose tissue and normally involves the intra-abdominal (increased visceral adipose tissue), breast, anterior neck, and dorsocervical regions.1 For this study, Model 2 was developed to assess fat in the trunk area – from the top of the iliac crest to the base of the skull – using subscapular skinfold which is highly correlated with DXA values (Table 2). This assessment is important since the accumulation of trunk fat is commonly associated with dyslipidemia, insulin resistance,2 and high intima-media carotid thickness17 in children and adolescents living with HIV and can lead to poor adherence to and failure of treatment.

Previous studies have reported correlations between anthropometric equations and DXA in healthy children and adolescents.25,26,28,29 Although DXA is a reproducible and accurate technique for the measurement of body fat, Silva et al.10 alerted to the development of equations using a multi-compartment approach in pediatric studies because of individual variability in the composition of fat-free mass. Moreover, the validity of the response variable is determinant for the development of appropriate equations based on anthropometry.10 Most validity studies of anthropometric equations in healthy children and adolescents have used DXA as the reference method.10 DXA was also used as the reference method to assess body fat in this study. Advantages of this method are its three-compartment approach, good precision, wide availability, and low radiation dose. However, DXA systems are not interchangeable and generalization is not possible, thus differences in developing predictive equations might be due to the use of different densitometers, softwares and/or scan modes.10

The equations of Lohman11 and Slaughter12 were not valid for estimating body fat in pediatric patients with HIV and their performance was lower than that of the equations developed in this study. These differences may be due to the reference methods used since the equations of Slaughter and Lohman were developed based on hydrostatic weighing, a densitometric method. However, in a study conducted on healthy Chinese children and adolescents with a similar age profile as the participants of the present study, the equation of Lohman was found to be a good predictor of body fat.28 In contrast, Mast et al.30 and Freedman et al.29 reported the equations of Lohman and Slaughter, respectively, to overestimate body fat in healthy children and adolescents compared to DXA.

Since limitations of the Lohman and Slaughter equations were found for the healthy population, some disagreement is expected in HIV-infected individuals due to body fat abnormalities. Children and adolescents living with HIV infected by mother-to-child transmission exhibit a delay in sexual development and low height and weight for age31 resulting from dysfunction of the hypothalamic-pituitary-adrenal axis.32 Evidently, the variability in growth and development will significantly change the body composition expected for age and render the traditional body fat equations impractical in this population. Therefore, we suggest the use of Models 1 and 2 to assess total body fat and trunk fat, respectively. Usually the validation process uses 2/3 of the sample for developing the prediction equation and 1/3 for cross-validation of the model. This was not possible in our study because of the small size and we have therefore used a bootstrap procedure of 5000 simulations for internal validation.

Some limitations suggest caution in the interpretation of the data obtained with regression models, such as the fact that the sample was from only one Brazilian region and the size of the sample used as reference. The strengths of the present study include the high coefficients of determination of the models (80% and 85%) and the low SEE. Furthermore, the practicality of anthropometry to monitor total body and trunk fat in the clinical setting was considered in the development of the models. The models were developed for children and adolescents living with HIV using a heterogeneous sample in terms of protease inhibitor use, clinical and immunological symptoms, and viral load, which can represent many limited-resources scenarios.

The models developed in this study for predicting total body and trunk fat in children and adolescents living with HIV based on anthropometric and demographic variables showed a high predictive power and agreed with the DXA method as demonstrated by an SEE of virtually zero. The traditional equations of Lohman and Slaughter showed poor performance and are not indicated for body fat assessment in pediatric HIV infection. Our prediction models (1 and 2) can be useful to monitor body fat in patients living with HIV, especially under conditions of limited resources because anthropometry is a noninvasive, low-cost and easy-to-use method.

Luiz R. A. de Lima conceived and designed the study, coordinated the data collection, interpreted the data, and wrote the manuscript. Priscila C. Martins, João A. C. de Castro and Carlos A. S. Alves Junior wrote the manuscript and analyzed and interpreted the data. Diego A. S. Silva and Edio L. Petroski interpreted the data and critically revised the study. All authors read and approved the final version of the manuscript.

The authors declare no conflicts of interest.