- Email: [email protected]
Prediction of Areal Bone Mineral Density and Bone Mineral Content in Children and Adolescents Living With HIV Based on Anthropometric Variables
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health, vol. -, no. -, 1e8, 2015 Ó Copyright 2015 by The International Society for Clinical Densitometry 1094-6950/-:1e8/$36.00 http://dx.doi.org/10.1016/j.jocd.2015.06.001
Original Article
Prediction of Areal Bone Mineral Density and Bone Mineral Content in Children and Adolescents Living With HIV Based on Anthropometric Variables Luiz Rodrigo Augustemak de Lima,*,1 Rodrigo de Rosso Krug,2 Rosane Carla Rosendo da Silva,1 Aroldo Prohmann de Carvalho,3,4,5 David Alejandro Gonz alez-Chica,6 Isabela de Carlos Back,7 and Edio Luiz Petroski1 1
Graduate Program in Physical Education, Sports Centre, Federal University of Santa Catarina, Florianopolis, Santa Catarina, Brazil; 2Graduate Program in Medical Sciences, Health Sciences Centre, Federal University of Santa Catarina, Florian opolis, Santa Catarina, Brazil; 3Pediatric Infectologist at Hospital Infantil Joana de Gusm~ao, Florianopolis, Santa Catarina, Brazil; 4Federal University of Santa Catarina, Florianopolis, Santa Catarina, Brazil; 5Vale do Itajaı University, Florian opolis, Santa Catarina, Brazil; 6Discipline of General Practice, School of Medicine, The University of Adelaide, Adelaide, South Australia, Australia; and 7Graduate Program in Public Health, Federal University of Santa Catarina, Health Sciences Centre, Florianopolis, Santa Catarina, Brazil
Abstract Children and adolescents living with HIV have low bone mass for age. There are reliable and accurate methods for evaluation of bone mass, however, alternative methods are necessary, especially, for application in limitedresource scenarios. Anthropometry is a noninvasive and low cost method that can predict bone mass in healthy youths. The aim of the study was to develop predictive equations for bone mineral content and bone mineral density in children and adolescents living with HIV based on anthropometric variables. Forty-eight children and adolescents of both sexes (24 females) from 7 to 17 years, living in greater Florianopolis area, Santa Catarina, Brazil, who were under clinical follow-up at ‘‘Hospital Infantil Joana de Gusm~ao’’, participated in the study. Dual-energy X-ray absorptiometry was used to evaluate whole-body bone mineral content (BMC) and areal bone mineral density (aBMD). Height, body weight, bone diameters, arm circumference, and triceps skinfold were measured and the body mass index and arm muscle area were calculated. Multiple regression models were fitted to predict BMC and aBMD, using backward selection ( p 0.05). Two predictive models with high R2 values (84%e94%) were developed. Model 1 to estimate aBMD [Y 5 0.1450124 þ (height 0.0033807) þ (age 0.0146381) þ (body mass index 0.0158838) þ (skin color 0.0421068)], and model 2 to estimate BMC [Y 5 1095.1 þ (body weight 45.66973) þ (age 31.36516) þ (arm circumference 53.27204) þ (femoral diameter 9.594018)].The predictive models using anthropometry provided reliable estimates and can be useful to monitor aBMD and BMC in children and adolescents living with human immunodeficiency virus where limited resources are available. Key Words: Acquired immunodeficiency syndrome; anthropometry; bone density; HIV; osteoporosis.
Introduction
Received 04/29/15; Revised 05/28/15; Accepted 06/17/15. *Address correspondence to: Luiz Rodrigo Augustemak de Lima, MSc, Universidade Federal de Santa Catarina, Campus Universitario Reitor Jo~ao David Ferreira Lima, Centro de Desportos, N ucleo de Pesquisa em Cineantropometria e Desempenho Humano, Bairro Trindade, CEP: 88040900. Florianopolis, Santa Catarina, Brazil. E-mail: [email protected]
Human immunodeficiency virus (HIV) itself and highly active antiretroviral therapy (HAART) are associated with adverse morphological, metabolic, cardiovascular, nervous system, and renal events (1). The available information suggests low bone mineral content (BMC) and areal bone 1
2 mineral density (aBMD) in children and adolescents infected with HIV (2). The main concern regarding seropositive individuals infected by vertical transmission are the long-term effects of exposure to HIV and to the drugs used for its treatment, particularly during puberty, a critical period of bone mineral accrual when about 80% of the adult bone mass is deposited. Low peak bone mass during this phase is related to an increased risk of osteoporosis and fractures (3). Bone mass can be measured by quantitative computed tomography, quantitative ultrasound, and dual-energy X-ray absorptiometry (DXA) (4). These methods are usually used to assess BMC, BMD, and aBMD, but they require sophisticated equipment, adequate physical structure, and are expensive in most of settings where HIV is more prevalent. Anthropometry could be alternative method for the predicting bone mass because it is noninvasive and of low cost. Body weight, height, circumference measurements, and bone diameters, combined with age, gender, and skin color, have been used for the prediction of bone mass in healthy children and adolescents (5). These variables can be measured with portable devices which are frequently available in primary care centers, although trained personnel are necessary for reliable measurement. In this respect, predictive equations would permit valid and accurate data that could be applied in clinical practice and in field studies (6). These predictive equations are particularly relevant to middle- and low-income countries where resources for bone health monitoring are limited. Despite these advantages, there are few studies designed to elaborate prediction models of bone mass for populations with altered clinical conditions (7,8). Therefore, the objective of the present study was to develop predictive equations of BMC and aBMD based on anthropometric measures in children and adolescents living with HIV using DXA as the reference method.
Methodology Study Design and Patient Population A cross-sectional and observational study was conducted in 2009 in the city of Florian opolis, capital of the State of Santa Catarina, Southern Brazil. The city has the third highest Human Development Index (0.847) in Brazil (average of 0.727). Additionally, the Gini coefficient, which is a measure of inequality of income distribution, is 0.566 in Florianopolis. In Brazil, the coefficient ranges from 0.284 to 0.808 (9). The target population consisted of children and adolescents living with HIV acquired by vertical transmission, who were under clinical follow-up at the specialized care service of ‘‘Hospital Infantil Joana de Gusm~ao’’ (HIJG). This hospital is a referral center of pediatric HIV treatment in the state. All patients seen between August and December 2009 were selected based on a screening performed in 2008 and participation in a previous study (10). Criteria for inclusion in the study were: HIV infection acquired by vertical
Lima et al. transmission; age between 7 and 17 years; clinical and laboratory records of HIV infection; regular treatment at HIJG; absence of concomitant diseases (e.g., renal and liver failure, hypothyroidism and hyperthyroidism, cancer) and no use of diuretic agents that would alter body composition, except for HIV antiretroviral therapy. Eighty-three children and adolescents infected with HIV were eligible for this study. However, 4 (2%) participants were excluded because of unavailability and 4 (2%) because of the lack of interest in participating in the study. The latter were considered refusals. Additionally, 9 (10.8%) participants were considered losses because of the absence on some of the assessments, 15 (18.0%) because they could not be contacted, and 3 (3.6%) because they did not undergo bone mass evaluation. Thus, the final sample consisted of 48 children and adolescents (24 of each sex). The study was conducted in accordance with the ethical principles of the Declaration of Helsinki and was approved by the Ethics Committee of HIJG (Protocol No. 077/2009). The parent or legal guardian signed the informed consent.
Dual-energy X-ray Absorptiometry Bone mass was evaluated by DXA using Hologic’s Discovery WI Fan-Bean system (Bedford, MA, USA) and X-ray attenuation was computed using a pediatric software (version 14.4:5). This method is safe because the equipment emits radiation of 4.2 to 5.2 mSv, corresponding to that received on a sunny day. DXA shows high validity, reproducibility, and accuracy in the measurement of bone mass (11). The device was daily calibrated according to the manual of the manufacturer. Whole-body phantoms were used monthly for calibration, guaranteeing internal quality control of the equipment. The average coefficient of variation of the equipment during the study was 1% for aBMD evaluation, a value similar to that reported in another study (11). During the assessments, the participants wore appropriate clothing without any type of metal and were barefoot. The measurements were standardized and performed by 2 radiology professionals at a specialized clinic. After total aBMD and BMC data were obtained, age-, gender-, and ethnicity-specific Z-scores were calculated based on LMS value (12) using the LMSgrowth Microsoft Excel add-in (South Shields, Tyne & Wear, UK) software (13). A more detailed information about these Z-score values was presented elsewhere (14).
Anthropometric Variables Height was measured with a Tonelli stadiometer (120A; Criciuma, Brazil) to the nearest 1 mm, and body weight was measured with a Tanita digital scale (BF683W, Arlington Heights, IL, USA) to the nearest 0.1 kg using standardized procedures (15). These data were used to calculate the body mass index (BMI) (16). Humeral and femoral bone diameters were measured with a digital caliper (Digimess, S~ao Paulo, Brazil) to the nearest 0.01 mm (15). Arm circumference was measured with a nonelastic anthropometric tape with a retraction spring to the nearest 0.1 cm (15). The
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015
Prediction of Bone Mass in HIV + Youths triceps skinfold was measured with a Cescorf (Cescorf, Porto Alegre, Brazil) skinfold caliper to the nearest 0.1 mm according to a standardized procedure (15). Arm circumference and triceps skinfold were used to calculate the upper arm muscle area (UAMA) (17). Body weight and height were measured in duplicate, while bone diameters, arm circumference, and skinfold thickness were obtained in triplicate. The means of the anthropometric variables were used for analyses. All anthropometric measurements were taken individually in a private room. The examiners were previously trained. The technical errors of measurement were established by evaluating a group of 16 healthy subjects with similar characteristics and were 0.19 cm and 0.51 kg for height and body weight, respectively, 0.55 mm for femoral diameter, 0.22 cm for arm circumference, and 0.41 mm for triceps skinfold. The intraclass correlation coefficient for the anthropometric measures ranged from 0.96 to 1.00.
Confounding Variables Confounding variables were evaluated considering that these variables can increase the predictive power of the equations to be developed. Information about age (in years), gender (male or female), and skin color (white or mulatto/black) of the children and adolescents participating in the study, as well as the socioeconomic level (number of minimum wages received) and education level (in years) of the parent/guardian were collected using a structured questionnaire applied by interview. Bone age was evaluated by X-ray of the left hand wrist at the radiology service of HIJG and calculated by a radiologist using the GreulichePyle method (18). Maturity status was determined by self-assessment of secondary sex characteristics after the participants had received instructions about the morphological alterations of each stage from a researcher of the same sex (19). From the medical records, clinical and laboratory variables such as immunological and clinical symptom classification, use of HAART, HIV RNA viral load, and CD4þ T lymphocyte were obtained. The last 2 were measured by branched-DNA assays and flow cytometry, respectively. The clinical evolution was classified as (1) asymptomatic or presence of mild symptoms and (2) presence of moderate and severe symptoms (20,21).
Statistical Analysis Symmetry and kurtosis were analyzed using graphs and the ShapiroeWilk test. Asymmetric data were normalized (1/Ovariable). Bone age were missing for one subject; thus, the hot deck imputation procedure (22) was applied considering sex, age, BMI, and skin color. Median and interquartile range (p25th; p75th) were used to describe the participants. The independent Student’s t test or ManneWhitney U-test was used to compare sexes. A correlation matrix was used to identify predictors of bone mass using Pearson’s linear correlation coefficient and Spearman rank correlation to categorical variables. The linearity was analyzed graphically. Predictive variables showing p ! 0.05 were selected for multiple linear regression analysis.
3 Modeling was performed using a backward selection procedure, removing variables with lower statistical significance ( p 0.05) in decreasing order to until the final model was obtained to predict BMC or aBMD. Both models were adjusted by sex, to take in account the sexual dimorphism effect. The models were evaluated considering the coefficient of determination, regression coefficients, and maximum likelihood test. Additionally, the practicality of obtaining the anthropometric variables and particular features of HIV, such as possible delays in biological maturation and the impact on growth, were considered. The mean estimated values and those measured by DXA were compared by the paired Student’s t test. Lin’s concordance correlation coefficient was used to test the association between data. Models were analyzed by the residuals, standard error of the estimate, and multicollinearity of the predictive variables, in addition to the analysis of tolerance or variance inflation factor, Akaike’s information criterion, and Bayesian information criterion. The BreuschePagan or CookeWeisberg test was used to evaluate the heteroskedasticity of residuals. The dispersion of residuals was verified using the BlandeAltman method. Bootstrap validation with 1000 bootstrap samples was used to examine the internal validity of develop models. Statistical analyses were performed using the STATA 11.0 (Stata Corporation, College Station, TX, USA) and GraphPad Prism 5.0 (GraphPad Software, Inc., San Diego, CA, USA) packages.
Results Table 1 shows the characteristics of the children and adolescents living with HIV. Twenty-seven (56.2%) participants were white and the families of 31 (64.6%) had a monthly income of !3 minimum wages (minimum wage 5 US $299.58). Fortytwo (88.2%) participants were eutrophic according to BMI for age. A delay in skeletal maturation was observed in five (10.4%) subjects, considering a difference 2.0 standard deviations for gender and age (18). Analysis of maturity revealed 7 (14.6%) in Tanner stage I, 15 (31.3%) were in stage II, 13 (27.1) in stage III, and 11 (22.9%) in stage IV, in addition, 2 (4.2%) are mature. The difference between bone and chronological age supports the delay in skeletal maturity. Negative median BMC and aBMD Z-scores were observed, indicating a low bone mass for age. The median time of HAART use was 9.1 years (range: 1.2e13.8 years); 54.2% of the subjects used protease inhibitors. In general, the clinical condition of the subjects was stable, with an undetectable viral load in more than half (n 5 28). The median CD4þ T lymphocyte count was 747.5 cells per mm3 (range: 233e2034 cells per mm3). However, most participants exhibited immunosuppression and moderate or severe clinical symptoms. Because no difference between sexes was observed in the sample studied, the subsequent results from boys and girls were pooled. Table 2 shows the correlation between predictive and response variables. Body weight and height showed
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015
4
Lima et al. Table 1 Characteristics of the 48 HIV-positive Children and Adolescents in a Study on the Prediction of Areal Bone Mineral Density and Bone Mineral Content Based on Anthropometric Variables (Florianopolis, Brazil, 2011) Boys (n 5 24)
Variable Chronological age (yr) Bone age (yr) Body weight (kg) Height (cm) BMI (kg m2) UAMA (mm2) Arm circumference (cm) Humeral diameter (mm) Femoral diameter (mm) Total aBMD (g cm2) Total BMC (g) z-BMD (standard deviations) z-BMC (standard deviations) Tanner stages I II III IV V Skin color White Black/mulatto Clinical symptoms Asymptomatic/mild Moderate/severe Immunosuppression Absent Moderate Severe Type of HAART With PI Without PI
Median 13.2 12.8 40.9 150.1 17.64 2701.7 21.05 61.66* 90.01* 0.902 1329.8 0.84 1.01
p25th; p75th 11.1; 14.7 10.0; 14.0 30.4; 46.9 138.6; 160.5 15.94; 18.82 2099.3; 3562.1 18.62; 23.77 56.2; 65.2 82.6; 93.2 0.831; 0.960 1088.9; 1616.8 1.66; 0.33 1.51; 0.44 n (%)
Girls (n 5 24) Median 12.7 12.5 39.3 148.6 17.98 2408.8 20.45 57.96 83.14 0.849 1227.6 1.28 1.07
p25th; p75th 9.4; 15.0 9.5; 15.5 30.4; 48.0 133.3; 153.2 16.47; 20.43 2175.7; 3367.1 19.52; 25.00 54.88; 59.63 78.47; 87.69 0.769; 0.968 1026.0; 1585.3 1.94; 0.42 1.42; 0.06 n (%)
2 (8.3) 10 (41.6) 7 (29.1) 4 (16.6) 1 (4.1)
5 (20.8) 5 (20.8) 6 (25.0) 7 (29.1) 1 (4.1)
13 (54.2) 11 (45.8)
14 (58.3) 10 (41.7)
11 (45.8) 13 (54.2)
14 (58.3) 10 (41.7)
6 (25.0) 13 (54.2) 5 (20.8)
3 (12.5) 15 (62.5) 6 (25.0)
15 (62.5) 9 (37.5)
11 (45.8) 13 (54.2)
Abbr: aBMD, areal bone mineral density; BMC, bone mineral content; BMI, body mass index; z-BMC, bone mineral content z-score; z-BMD, bone mineral density z-score; HAART, highly active antiretroviral therapy; HIV, human immunodeficiency virus; PI, protease inhibitors; UAMA, upper arm muscle area. *p ! 0.05 between sexes.
the highest correlations with BMC and BMC, followed by age and 1/OUAMA. Table 3 shows the prediction models for evaluation of bone mass based on bivariate (Table 2) and multiple analyses. Model 1 explained 84% of aBMD measured by DXA based on height (standardized beta coefficient 5 0.32), BMI (0.33), age (0.31), skin color (0.15) as predictors, adjusted by sex; this model showed a low standard error of the estimate (SEE) (0.048 g). Model 2, which included
body weight (beta standard coefficient 5 1.24), arm circumference (0.33), femoral diameter (0.20), and age (0.19), adjusted by sex, had a high predictive power of BMC, explaining 94% of the variance, with a SEE of 109.3 g. Alternative models to predict aBMD and BMC, also total less head, either adjusted by maturity can be found in Supplementary Table 1. Table 4 shows no differences between aBMD or BMC measured by DXA with the estimates obtained with models
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015
Prediction of Bone Mass in HIV + Youths
5
Table 2 Correlation Between the Response Variables and Predictors Obtained for the 48 HIV-positive Children and Adolescents in a Study on the Prediction of Areal Bone Mineral Density and Bone Mineral Content Based on Anthropometric Variables (Florianopolis, Brazil, 2011) Total aBMD (g cm2)
Total BMC (g) Variable
Pearson’s correlation coefficient (r)
Body weight (kg) Height (cm) Age (yr) Bone age (yr) 1/OUAMA (mm2) Humeral diameter (mm) Arm circumference (cm) Femoral diameter (mm) BMI (kg m2) Sexual maturity (Tanner stage) Time of HAART (yr) Skin color (white; mulatto/black) Clinical symptoms (asymptomatic/mild; moderate/severe) Type of HAART (with PI; without PI)
0.95* 0.90* 0.84* 0.83* 0.81* 0.80* 0.78* 0.78* 0.76* 0.75*** 0.56* 0.29** 0.09 0.06
0.89* 0.85* 0.85* 0.81* 0.81* 0.75* 0.76* 0.75* 0.75* 0.75* 0.59* 0.36*** 0.14 0.08
Note: For categorical variables Spearman Rank Correlation was used. Abbr: aBMC, areal bone mineral content; BMD, bone mineral density; BMI, body mass index; HAART, combination antiretroviral therapy; HIV, human immunodeficiency virus; PI, protease inhibitors; UAMA, upper arm muscle area. *p ! 0.001; **p ! 0.05; ***p ! 0.01.
1 and 2 ( p O 0.05). Furthermore, Lin’s concordance correlation coefficients were of moderate to substantial magnitude. The linearity between the values measured by DXA and the estimates obtained with the models and also the analyses of
residuals are shown in Figs. 1 and 2, respectively. A random distribution of residuals for aBMD and BMC was observed for models 1 and 2, respectively (x2 5 3.05 and p 5 0.08; x2 5 1.60 and p 5 0.20).
Table 3 Multiple Regression Models for the Prediction of Bone Mass in Children and Adolescents Living With HIV (Florian opolis, Brazil, 2011) Multiple Regression Models Dependent variable Total aBMD (g cm2) Model 1* Y 5 0.1321904 þ (height 0.0028424) þ (age 0.0153856) þ (BMI 0.01837) þ (skin color 0.0412933) þ (sex 0.0273668) Dependent variable Total BMC (g) Model 2* Y 5 1193.955 þ (BW 45.70223) þ (age 32.50329) þ (AC 48.89363) þ (FD 12.30302) þ (sex 46.86491)
R2adjusted
0.84
0.94
SEE
0.048
109.28
RMSE
0.055
115.48
F
AIC n
BIC0
T
VIF
50.25
138.473
73.929
0.16
2.75
140.30
597.72
118.581
0.06
5.01
Note: Height in cm; age in years; skin color (1 5 white; 2 5 black/mulatto); sex (0 5 female, 1 5 male). Abbr: aBMD, areal bone mineral density; AC, arm circumference (cm); AIC, Akaike’s information criterion; BIC, Bayesian information criterion; BMC, bone mineral content; BMI, body mass index (kg m2); BW, body weight (kg); FD, femoral diameter (mm); HIV, human immunodeficiency virus; RMSE, root mean square error; SEE, standard error of the estimate; T, tolerance; VIF, variance inflation factor. *p ! 0.001. Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015
6
Lima et al. Table 4 Comparison Between Bone Mass Measured by DXA and Estimated With the Prediction Models (Models 1 and 2) in Children and Adolescents Living With HIV (Florianopolis, Brazil, 2011)
Measured and Estimated Bone Mass Total Total Total Total
aBMDDXA aBMDmodel 1 BMCDXA BMCmodel 2
Mean (SD) 0.886 0.886 1368.1 1368.1
(0.135) (0.125) (459.3) (446.1)
Rho ( p)
TE
AD
0.92 (!0.001)
0.007
!0.001
0.97 (!0.001)
!0.001
15.59
Note: There was no difference between the mean values obtained by DXA and those estimated with the models (t test 5 0.0000). Abbr: aBMD, areal bone mineral density; AD, average difference between measured and estimated values; BMC, bone mineral content; DXA, dual-energy X-ray absorptiometry; HIV, human immunodeficiency virus; Rho, Lin’s concordance correlation coefficient; SD, standard deviation; TE, total error.
Internal validation with 1000 bootstrap samples produced a corrected R2 of 84% and 94% to aBMD and BMC models, respectively.
Discussion The present study demonstrated that height, body weight, arm circumference, BMI, and femoral diameter, with age and skin color adjusted by sex, predict in 84% and 94% of aBMD and BMC, respectively, in children and adolescents living with HIV. These variables are accurate and reliable and can be easily measured in clinical setting. DXA, which is the preferred method for the evaluation of aBMD and BMC (23), was used as reference for the development of these models. These results are relevant to children and adolescents vertically infected with HIV living in Brazil, but with subsequent validation these models could be useful for bone mass monitoring elsewhere. Databases of healthy subjects can provide the parameters for the calculation of Z-scores (12) and consequently, the clinical interpretation of BMC and aBMD for age.
Previous studies have reported anthropometric variables as potential predictors of bone mass in children, (8,24,25) adolescents, (5,26) adults, and the elderly (7). As a previous research (24), we founded that height was the best predictor of aBMD in children and adolescents of both sexes. However, in a study with adolescent girls (26) fat-free mass presented the highest correlation with aBMD at different anatomical sites and was also the best predictor of aBMD in other study explaining 26% of bone mass (25). However, body weight was an important predictor of aBMD in children and young adults with cerebral palsy (8) and the correlations of body weight, height, BMI, and fat-free mass with aBMD were low to moderate in other study (25). We found that BMI, height, and age had a similar predictive power of aBMD, while body weight was the best predictor of BMC, followed by arm circumference, femoral diameter, and age. Anthropometric measurements, when combined, can show excellent predictive power of bone mass, and this was shown in prepubertal subjects (24) when height, diameters, skinfolds, and circumferences explained 60%e91% of aBMD and BMC. Our study corroborates these findings. The models explained
Fig. 1. Relationship between bone mass measured by DXA (aBMDDXA and BMCDXA) and predicted by the equations developed (aBMDmodel 1 and BMCmodel 2) in children and adolescents living with HIV (Florianopolis, Brazil, 2011). aBMD, areal bone mineral density; BMC, bone mineral content; DXA, dual-energy X-ray absorptiometry; HIV, human immunodeficiency virus. Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015
Prediction of Bone Mass in HIV + Youths
7
Fig. 2. Analysis of residual scores of areal bone mineral density (A) and bone mineral content (B) measured by DXA and estimated with the prediction models in children and adolescents living with HIV (Florianopolis, Brazil, 2011). aBMD, areal bone mineral density; BMC, bone mineral content; DXA, dual-energy X-ray absorptiometry; HIV, human immunodeficiency virus. from 84% to 94% of aBMD and BMC variance, respectively; although most participants were adolescents. This fact would imply more variability in bone mass, which is also affected by a delay in growth and maturity in HIV (27). In a Brazilian study, on the prediction of aBMD in adolescents (25), sex, fat-free mass, and abdominal muscle resistance explained approximately 40% of aBMD. In stratified analysis, abdominal resistance and maximum oxygen uptake and fat-free mass were the best predictors in boys and girls, respectively. Moreover, in North American adolescents, leg strength and power were predictors of aBMD and BMC (26). Both studies demonstrated the dynamic interaction between morphology and function and the possible prediction by physical fitness. However, several studies developing predictive or explanatory models of bone mass have considered many variables that represent size, such as diameters and height, in addition to others that can be proxies for gravitational or tension overload such as body weight, arm circumference, and UAMA (8,24e26). This is in agreement with the theoretical assumption and biological plausibility of the measurement of areal aBMD by DXA (5,28) and of osteogenic stimuli to induce bone formation (29), respectively. Body measures and the protein content of fat-free mass increase with age (30), a fact that would explain their importance in prediction models. Moreover, black or mulatto skin color has been shown to be associated with higher BMC compared with white skin color (12). Moreover, the models accounted for the sexual dimorphism in bone mass (31). We recommend the use of both models for the estimation of BMC and aBMD. The analyses of linearity highlight the high predictive power of the 2 models. Furthermore, correlation coefficients of moderate to substantial magnitude were obtained. The residuals showed a random distribution of the estimated values, while the average difference between measured and estimated values was virtually zero. The bootstrap procedure guaranteed an internal validity. Model 1 adequately estimated aBMD, from 100 to þ100 g cm2, in 97.9% of the children and adolescents studied. In a
restricted range, from 50 to þ50 g cm2, approximately 70% of the participants presented accurate estimates of aBMD. The same was observed for the BMC values with model 2 accurately estimating (200 to þ200 g) BMC in 93.7% of the participants. Within a restricted range, 100 to þ 100 g, 66.6% of the participants had valid estimates. However, the literature indicates the use of BMC due to the greater reproducibility and the absence of errors compared to areal density (23,28). Despite some limitations that suggest caution in interpreting the data obtained with the regression models, such as regional Brazilian sample as reference. The strengths of the present study include the high coefficient of determination (84%e94%) and the low standard errors of the estimate. Furthermore, to develop the regression models was considered the practicality of clinical application and the analysis of statistical parameters for robustness and reliability of the estimates. The models were developed for children and adolescents living with HIV using a heterogeneous sample in terms of use of protease inhibitors, clinical and immunological symptoms, and viral load. In conclusion, the present study developed prediction models of BMC and aBMD for children and adolescents living with HIV based on anthropometric and demographic variables. The models show high predictive power and agreed with the DXA method as the SEE are virtually zero. These prediction models can be useful to monitor bone mass in patients living with HIV, especially in scenarios in which resources are limited because anthropometry is easily available.
Acknowledgment The authors acknowledge the patients, investigators, and study personnel participating in this study. Also, we would like to thank Coordination of Improvement of Higher Education Personal (CAPES) for the awarded scholarships and Clınica Imagem for the assistance with DXA scans.
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015
8
Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jocd.2015.06.001.
References 1. Hazra R, Siberry GK, Mofenson LM. 2010 Growing up with HIV: children, adolescents, and young adults with perinatally acquired HIV infection. Annu Rev Med 61:169e185. 2. Arpadi SM, Shiau S, Arpadi CM, Yin MT. 2014 Bone health in HIV-infected children, adolescents and young adults: a systematic review. J AIDS Clin Res 5:2. 3. Rizzoli R, Bonjour JP. 1999 Determinants of peak bone mass and mechanisms of bone loss. Osteoporos Int 9(Suppl 2): S17eS23. 4. Specker BL, Schoenau E. 2005 Quantitative bone analysis in children: current methods and recommendations. J Pediatr 146:726e731. 5. Katzman DK, Bachrach LK, Carter DR, Marcus R. 1991 Clinical and anthropometric correlates of bone mineral acquisition in healthy adolescent girls. J Clin Endocrinol Metab 73:1332e1339. 6. Belissari A, Roche AF. 2005 Anthropometry and ultrasound. In: Human Body Composition. Heymsfield SB, Lohman TG, Wang Z, Going SB, eds. 2a ed. Champaign: Human Kinetics, 109e127. 7. Casini A, Mohamed EI, Gandin C, et al. 2003 Predicting bone mineral density of postmenopausal healthy and cirrhotic Italian women using anthropometric variables. Dig Liver Dis 35:881e887. 8. Henderson RC, Kairalla J, Abbas A, Stevenson RD. 2004 Predicting low bone density in children and young adults with quadriplegic cerebral palsy. Dev Med Child Neurol 46:416e419. 9. United Nations Development Programme. Institute for Applied Economic Research. Jo~ao Pinheiro Foundation. Atlas of Human Development in Brazil. Available at: http://www.atlasbrasil.org. br/2013/pt/. Accessed: November 20, 2014. 10. Giuliano IDCB, De Freitas SFT, De Souza M, Caramelli B. 2008 Subclinic atherosclerosis and cardiovascular risk factors in HIVinfected children: PERI study. Coron Artery Dis 19:167e172. 11. Leonard CM, Roza MA, Barr RD, Webber CE. 2009 Reproducibility of DXA measurements of bone mineral density and body composition in children. Pediatr Radiol 39:148e154. 12. Kelly TL, Wilson KE, Heymsfield SB. 2009 Dual energy X-ray absorptiometry body composition reference values from NHANES. PLoS One 4:e7038. 13. Pan H, Cole TJ. LMSgrowth, a Microsoft Excel add-in to access growth references based on the LMS method. Version 2.77. http:// www.healthforallchildren.co.uk/; 2012. Accessed: September 12, 2011. 14. Lima LRA, Silva RCR, Giuliano ICB, et al. 2013 Bone mass in children and adolescents infected with human immunodeficiency virus. J Pediatr 1:91e99. 15. Lohman TG, Roche AF, Martorell R. 1991 Anthropometric Standardization Reference Manual. Abridged ed. Champaign: Human Kinetics.
Lima et al. 16. Onis M, Onyango AW, Borghi E, et al. 2007 Development of a WHO growth reference for school-aged children and adolescents. Bull World Health Organ 85:660e667. 17. Frisancho AR. 1981 New norms of upper limb fat and muscle areas for assessment of nutritional status. Am J Clin Nutr 34: 2540e2545. 18. Greulich WW, Pyle SI. 1959 Radiographic atlas of skeletal development of the hand and wrist. Stanford, California: Stanford University Press. 19. Tanner JM. 1962 Growth at adolescence. 2nd ed. Oxford: Blackwell Scientific Publications. 20. Centers for Disease Control and Prevention. 1994 Revised classification system for human immunodeficiency virus infection in children less than 13 years of age; Official authorized addenda: human immunodeficiency virus infection codes and official guidelines for coding and reporting ICD-9-CM. MMWR 1994; 43(No. RR-12):[1-10]. 21. Centers for Disease Control and Prevention. 2008 Revised surveillance case definitions for HIV infection among adults, adolescents, and children aged !18 months and for HIV infection and AIDS among children aged 18 months to !13 yearsd United States. MMWR Recomm Rep 57(No. RR-10):1e12. 22. Haukoos JS, Newgard CD. 2007 Advanced statistics: missing data in clinical researchepart 1: an introduction and conceptual framework. Acad Emerg Med 14:662e668. 23. Crabtree NJ, Arabi A, Bachrach LK, et al. 2014 Dual-energy X-ray absorptiometry interpretation and reporting in children and adolescents: the revised 2013 ISCD Pediatric Official Positions. J Clin Densitom 17:225e242. 24. Miller JZ, Slemenda CW, Meaney FJ, et al. 1991 The relationship of bone mineral density and anthropometric variables in healthy male and female children. Bone Miner 14:137e152. 25. Fonseca RM, de Franca NM, Van Praagh E. 2008 Relationship between indicators of fitness and bone density in adolescent Brazilian children. Pediatr Exerc Sci 20:40e49. 26. Witzke KA, Snow CM. 1999 Lean body mass and leg power best predict bone mineral density in adolescent girls. Med Sci Sports Exerc 31:1558e1563. 27. de Martino M, Tovo PA, Galli L, et al. 2001 Puberty in perinatal HIV-1 infection: a multicentre longitudinal study of 212 children. AIDS 15:1527e1534. 28. Prentice A, Parsons TJ, Cole TJ. 1994 Uncritical use of bone mineral density in absorptiometry may lead to size-related artifacts in the identification of bone mineral determinants. Am J Clin Nutr 60:837e842. 29. Judex S, Carlson KJ. 2009 Is bone’s response to mechanical signals dominated by gravitational loading? Med Sci Sports Exerc 41:2037e2043. 30. Malina RM, Bouchard C, Bar-Or O. 2004 Growth, maturation, and physical activity. Champaign, Illinois: Human Kinetics. 31. Lu J, Shin Y, Yen MS, Sun SS. [In Press, available online 18 October 2014] Peak bone mass and patterns of change in total bone mineral density and bone mineral contents from childhood into young adulthood. J Clin Densitom.
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015
Original Article
Prediction of Areal Bone Mineral Density and Bone Mineral Content in Children and Adolescents Living With HIV Based on Anthropometric Variables Luiz Rodrigo Augustemak de Lima,*,1 Rodrigo de Rosso Krug,2 Rosane Carla Rosendo da Silva,1 Aroldo Prohmann de Carvalho,3,4,5 David Alejandro Gonz alez-Chica,6 Isabela de Carlos Back,7 and Edio Luiz Petroski1 1
Graduate Program in Physical Education, Sports Centre, Federal University of Santa Catarina, Florianopolis, Santa Catarina, Brazil; 2Graduate Program in Medical Sciences, Health Sciences Centre, Federal University of Santa Catarina, Florian opolis, Santa Catarina, Brazil; 3Pediatric Infectologist at Hospital Infantil Joana de Gusm~ao, Florianopolis, Santa Catarina, Brazil; 4Federal University of Santa Catarina, Florianopolis, Santa Catarina, Brazil; 5Vale do Itajaı University, Florian opolis, Santa Catarina, Brazil; 6Discipline of General Practice, School of Medicine, The University of Adelaide, Adelaide, South Australia, Australia; and 7Graduate Program in Public Health, Federal University of Santa Catarina, Health Sciences Centre, Florianopolis, Santa Catarina, Brazil
Abstract Children and adolescents living with HIV have low bone mass for age. There are reliable and accurate methods for evaluation of bone mass, however, alternative methods are necessary, especially, for application in limitedresource scenarios. Anthropometry is a noninvasive and low cost method that can predict bone mass in healthy youths. The aim of the study was to develop predictive equations for bone mineral content and bone mineral density in children and adolescents living with HIV based on anthropometric variables. Forty-eight children and adolescents of both sexes (24 females) from 7 to 17 years, living in greater Florianopolis area, Santa Catarina, Brazil, who were under clinical follow-up at ‘‘Hospital Infantil Joana de Gusm~ao’’, participated in the study. Dual-energy X-ray absorptiometry was used to evaluate whole-body bone mineral content (BMC) and areal bone mineral density (aBMD). Height, body weight, bone diameters, arm circumference, and triceps skinfold were measured and the body mass index and arm muscle area were calculated. Multiple regression models were fitted to predict BMC and aBMD, using backward selection ( p 0.05). Two predictive models with high R2 values (84%e94%) were developed. Model 1 to estimate aBMD [Y 5 0.1450124 þ (height 0.0033807) þ (age 0.0146381) þ (body mass index 0.0158838) þ (skin color 0.0421068)], and model 2 to estimate BMC [Y 5 1095.1 þ (body weight 45.66973) þ (age 31.36516) þ (arm circumference 53.27204) þ (femoral diameter 9.594018)].The predictive models using anthropometry provided reliable estimates and can be useful to monitor aBMD and BMC in children and adolescents living with human immunodeficiency virus where limited resources are available. Key Words: Acquired immunodeficiency syndrome; anthropometry; bone density; HIV; osteoporosis.
Introduction
Received 04/29/15; Revised 05/28/15; Accepted 06/17/15. *Address correspondence to: Luiz Rodrigo Augustemak de Lima, MSc, Universidade Federal de Santa Catarina, Campus Universitario Reitor Jo~ao David Ferreira Lima, Centro de Desportos, N ucleo de Pesquisa em Cineantropometria e Desempenho Humano, Bairro Trindade, CEP: 88040900. Florianopolis, Santa Catarina, Brazil. E-mail: [email protected]
Human immunodeficiency virus (HIV) itself and highly active antiretroviral therapy (HAART) are associated with adverse morphological, metabolic, cardiovascular, nervous system, and renal events (1). The available information suggests low bone mineral content (BMC) and areal bone 1
2 mineral density (aBMD) in children and adolescents infected with HIV (2). The main concern regarding seropositive individuals infected by vertical transmission are the long-term effects of exposure to HIV and to the drugs used for its treatment, particularly during puberty, a critical period of bone mineral accrual when about 80% of the adult bone mass is deposited. Low peak bone mass during this phase is related to an increased risk of osteoporosis and fractures (3). Bone mass can be measured by quantitative computed tomography, quantitative ultrasound, and dual-energy X-ray absorptiometry (DXA) (4). These methods are usually used to assess BMC, BMD, and aBMD, but they require sophisticated equipment, adequate physical structure, and are expensive in most of settings where HIV is more prevalent. Anthropometry could be alternative method for the predicting bone mass because it is noninvasive and of low cost. Body weight, height, circumference measurements, and bone diameters, combined with age, gender, and skin color, have been used for the prediction of bone mass in healthy children and adolescents (5). These variables can be measured with portable devices which are frequently available in primary care centers, although trained personnel are necessary for reliable measurement. In this respect, predictive equations would permit valid and accurate data that could be applied in clinical practice and in field studies (6). These predictive equations are particularly relevant to middle- and low-income countries where resources for bone health monitoring are limited. Despite these advantages, there are few studies designed to elaborate prediction models of bone mass for populations with altered clinical conditions (7,8). Therefore, the objective of the present study was to develop predictive equations of BMC and aBMD based on anthropometric measures in children and adolescents living with HIV using DXA as the reference method.
Methodology Study Design and Patient Population A cross-sectional and observational study was conducted in 2009 in the city of Florian opolis, capital of the State of Santa Catarina, Southern Brazil. The city has the third highest Human Development Index (0.847) in Brazil (average of 0.727). Additionally, the Gini coefficient, which is a measure of inequality of income distribution, is 0.566 in Florianopolis. In Brazil, the coefficient ranges from 0.284 to 0.808 (9). The target population consisted of children and adolescents living with HIV acquired by vertical transmission, who were under clinical follow-up at the specialized care service of ‘‘Hospital Infantil Joana de Gusm~ao’’ (HIJG). This hospital is a referral center of pediatric HIV treatment in the state. All patients seen between August and December 2009 were selected based on a screening performed in 2008 and participation in a previous study (10). Criteria for inclusion in the study were: HIV infection acquired by vertical
Lima et al. transmission; age between 7 and 17 years; clinical and laboratory records of HIV infection; regular treatment at HIJG; absence of concomitant diseases (e.g., renal and liver failure, hypothyroidism and hyperthyroidism, cancer) and no use of diuretic agents that would alter body composition, except for HIV antiretroviral therapy. Eighty-three children and adolescents infected with HIV were eligible for this study. However, 4 (2%) participants were excluded because of unavailability and 4 (2%) because of the lack of interest in participating in the study. The latter were considered refusals. Additionally, 9 (10.8%) participants were considered losses because of the absence on some of the assessments, 15 (18.0%) because they could not be contacted, and 3 (3.6%) because they did not undergo bone mass evaluation. Thus, the final sample consisted of 48 children and adolescents (24 of each sex). The study was conducted in accordance with the ethical principles of the Declaration of Helsinki and was approved by the Ethics Committee of HIJG (Protocol No. 077/2009). The parent or legal guardian signed the informed consent.
Dual-energy X-ray Absorptiometry Bone mass was evaluated by DXA using Hologic’s Discovery WI Fan-Bean system (Bedford, MA, USA) and X-ray attenuation was computed using a pediatric software (version 14.4:5). This method is safe because the equipment emits radiation of 4.2 to 5.2 mSv, corresponding to that received on a sunny day. DXA shows high validity, reproducibility, and accuracy in the measurement of bone mass (11). The device was daily calibrated according to the manual of the manufacturer. Whole-body phantoms were used monthly for calibration, guaranteeing internal quality control of the equipment. The average coefficient of variation of the equipment during the study was 1% for aBMD evaluation, a value similar to that reported in another study (11). During the assessments, the participants wore appropriate clothing without any type of metal and were barefoot. The measurements were standardized and performed by 2 radiology professionals at a specialized clinic. After total aBMD and BMC data were obtained, age-, gender-, and ethnicity-specific Z-scores were calculated based on LMS value (12) using the LMSgrowth Microsoft Excel add-in (South Shields, Tyne & Wear, UK) software (13). A more detailed information about these Z-score values was presented elsewhere (14).
Anthropometric Variables Height was measured with a Tonelli stadiometer (120A; Criciuma, Brazil) to the nearest 1 mm, and body weight was measured with a Tanita digital scale (BF683W, Arlington Heights, IL, USA) to the nearest 0.1 kg using standardized procedures (15). These data were used to calculate the body mass index (BMI) (16). Humeral and femoral bone diameters were measured with a digital caliper (Digimess, S~ao Paulo, Brazil) to the nearest 0.01 mm (15). Arm circumference was measured with a nonelastic anthropometric tape with a retraction spring to the nearest 0.1 cm (15). The
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015
Prediction of Bone Mass in HIV + Youths triceps skinfold was measured with a Cescorf (Cescorf, Porto Alegre, Brazil) skinfold caliper to the nearest 0.1 mm according to a standardized procedure (15). Arm circumference and triceps skinfold were used to calculate the upper arm muscle area (UAMA) (17). Body weight and height were measured in duplicate, while bone diameters, arm circumference, and skinfold thickness were obtained in triplicate. The means of the anthropometric variables were used for analyses. All anthropometric measurements were taken individually in a private room. The examiners were previously trained. The technical errors of measurement were established by evaluating a group of 16 healthy subjects with similar characteristics and were 0.19 cm and 0.51 kg for height and body weight, respectively, 0.55 mm for femoral diameter, 0.22 cm for arm circumference, and 0.41 mm for triceps skinfold. The intraclass correlation coefficient for the anthropometric measures ranged from 0.96 to 1.00.
Confounding Variables Confounding variables were evaluated considering that these variables can increase the predictive power of the equations to be developed. Information about age (in years), gender (male or female), and skin color (white or mulatto/black) of the children and adolescents participating in the study, as well as the socioeconomic level (number of minimum wages received) and education level (in years) of the parent/guardian were collected using a structured questionnaire applied by interview. Bone age was evaluated by X-ray of the left hand wrist at the radiology service of HIJG and calculated by a radiologist using the GreulichePyle method (18). Maturity status was determined by self-assessment of secondary sex characteristics after the participants had received instructions about the morphological alterations of each stage from a researcher of the same sex (19). From the medical records, clinical and laboratory variables such as immunological and clinical symptom classification, use of HAART, HIV RNA viral load, and CD4þ T lymphocyte were obtained. The last 2 were measured by branched-DNA assays and flow cytometry, respectively. The clinical evolution was classified as (1) asymptomatic or presence of mild symptoms and (2) presence of moderate and severe symptoms (20,21).
Statistical Analysis Symmetry and kurtosis were analyzed using graphs and the ShapiroeWilk test. Asymmetric data were normalized (1/Ovariable). Bone age were missing for one subject; thus, the hot deck imputation procedure (22) was applied considering sex, age, BMI, and skin color. Median and interquartile range (p25th; p75th) were used to describe the participants. The independent Student’s t test or ManneWhitney U-test was used to compare sexes. A correlation matrix was used to identify predictors of bone mass using Pearson’s linear correlation coefficient and Spearman rank correlation to categorical variables. The linearity was analyzed graphically. Predictive variables showing p ! 0.05 were selected for multiple linear regression analysis.
3 Modeling was performed using a backward selection procedure, removing variables with lower statistical significance ( p 0.05) in decreasing order to until the final model was obtained to predict BMC or aBMD. Both models were adjusted by sex, to take in account the sexual dimorphism effect. The models were evaluated considering the coefficient of determination, regression coefficients, and maximum likelihood test. Additionally, the practicality of obtaining the anthropometric variables and particular features of HIV, such as possible delays in biological maturation and the impact on growth, were considered. The mean estimated values and those measured by DXA were compared by the paired Student’s t test. Lin’s concordance correlation coefficient was used to test the association between data. Models were analyzed by the residuals, standard error of the estimate, and multicollinearity of the predictive variables, in addition to the analysis of tolerance or variance inflation factor, Akaike’s information criterion, and Bayesian information criterion. The BreuschePagan or CookeWeisberg test was used to evaluate the heteroskedasticity of residuals. The dispersion of residuals was verified using the BlandeAltman method. Bootstrap validation with 1000 bootstrap samples was used to examine the internal validity of develop models. Statistical analyses were performed using the STATA 11.0 (Stata Corporation, College Station, TX, USA) and GraphPad Prism 5.0 (GraphPad Software, Inc., San Diego, CA, USA) packages.
Results Table 1 shows the characteristics of the children and adolescents living with HIV. Twenty-seven (56.2%) participants were white and the families of 31 (64.6%) had a monthly income of !3 minimum wages (minimum wage 5 US $299.58). Fortytwo (88.2%) participants were eutrophic according to BMI for age. A delay in skeletal maturation was observed in five (10.4%) subjects, considering a difference 2.0 standard deviations for gender and age (18). Analysis of maturity revealed 7 (14.6%) in Tanner stage I, 15 (31.3%) were in stage II, 13 (27.1) in stage III, and 11 (22.9%) in stage IV, in addition, 2 (4.2%) are mature. The difference between bone and chronological age supports the delay in skeletal maturity. Negative median BMC and aBMD Z-scores were observed, indicating a low bone mass for age. The median time of HAART use was 9.1 years (range: 1.2e13.8 years); 54.2% of the subjects used protease inhibitors. In general, the clinical condition of the subjects was stable, with an undetectable viral load in more than half (n 5 28). The median CD4þ T lymphocyte count was 747.5 cells per mm3 (range: 233e2034 cells per mm3). However, most participants exhibited immunosuppression and moderate or severe clinical symptoms. Because no difference between sexes was observed in the sample studied, the subsequent results from boys and girls were pooled. Table 2 shows the correlation between predictive and response variables. Body weight and height showed
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015
4
Lima et al. Table 1 Characteristics of the 48 HIV-positive Children and Adolescents in a Study on the Prediction of Areal Bone Mineral Density and Bone Mineral Content Based on Anthropometric Variables (Florianopolis, Brazil, 2011) Boys (n 5 24)
Variable Chronological age (yr) Bone age (yr) Body weight (kg) Height (cm) BMI (kg m2) UAMA (mm2) Arm circumference (cm) Humeral diameter (mm) Femoral diameter (mm) Total aBMD (g cm2) Total BMC (g) z-BMD (standard deviations) z-BMC (standard deviations) Tanner stages I II III IV V Skin color White Black/mulatto Clinical symptoms Asymptomatic/mild Moderate/severe Immunosuppression Absent Moderate Severe Type of HAART With PI Without PI
Median 13.2 12.8 40.9 150.1 17.64 2701.7 21.05 61.66* 90.01* 0.902 1329.8 0.84 1.01
p25th; p75th 11.1; 14.7 10.0; 14.0 30.4; 46.9 138.6; 160.5 15.94; 18.82 2099.3; 3562.1 18.62; 23.77 56.2; 65.2 82.6; 93.2 0.831; 0.960 1088.9; 1616.8 1.66; 0.33 1.51; 0.44 n (%)
Girls (n 5 24) Median 12.7 12.5 39.3 148.6 17.98 2408.8 20.45 57.96 83.14 0.849 1227.6 1.28 1.07
p25th; p75th 9.4; 15.0 9.5; 15.5 30.4; 48.0 133.3; 153.2 16.47; 20.43 2175.7; 3367.1 19.52; 25.00 54.88; 59.63 78.47; 87.69 0.769; 0.968 1026.0; 1585.3 1.94; 0.42 1.42; 0.06 n (%)
2 (8.3) 10 (41.6) 7 (29.1) 4 (16.6) 1 (4.1)
5 (20.8) 5 (20.8) 6 (25.0) 7 (29.1) 1 (4.1)
13 (54.2) 11 (45.8)
14 (58.3) 10 (41.7)
11 (45.8) 13 (54.2)
14 (58.3) 10 (41.7)
6 (25.0) 13 (54.2) 5 (20.8)
3 (12.5) 15 (62.5) 6 (25.0)
15 (62.5) 9 (37.5)
11 (45.8) 13 (54.2)
Abbr: aBMD, areal bone mineral density; BMC, bone mineral content; BMI, body mass index; z-BMC, bone mineral content z-score; z-BMD, bone mineral density z-score; HAART, highly active antiretroviral therapy; HIV, human immunodeficiency virus; PI, protease inhibitors; UAMA, upper arm muscle area. *p ! 0.05 between sexes.
the highest correlations with BMC and BMC, followed by age and 1/OUAMA. Table 3 shows the prediction models for evaluation of bone mass based on bivariate (Table 2) and multiple analyses. Model 1 explained 84% of aBMD measured by DXA based on height (standardized beta coefficient 5 0.32), BMI (0.33), age (0.31), skin color (0.15) as predictors, adjusted by sex; this model showed a low standard error of the estimate (SEE) (0.048 g). Model 2, which included
body weight (beta standard coefficient 5 1.24), arm circumference (0.33), femoral diameter (0.20), and age (0.19), adjusted by sex, had a high predictive power of BMC, explaining 94% of the variance, with a SEE of 109.3 g. Alternative models to predict aBMD and BMC, also total less head, either adjusted by maturity can be found in Supplementary Table 1. Table 4 shows no differences between aBMD or BMC measured by DXA with the estimates obtained with models
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015
Prediction of Bone Mass in HIV + Youths
5
Table 2 Correlation Between the Response Variables and Predictors Obtained for the 48 HIV-positive Children and Adolescents in a Study on the Prediction of Areal Bone Mineral Density and Bone Mineral Content Based on Anthropometric Variables (Florianopolis, Brazil, 2011) Total aBMD (g cm2)
Total BMC (g) Variable
Pearson’s correlation coefficient (r)
Body weight (kg) Height (cm) Age (yr) Bone age (yr) 1/OUAMA (mm2) Humeral diameter (mm) Arm circumference (cm) Femoral diameter (mm) BMI (kg m2) Sexual maturity (Tanner stage) Time of HAART (yr) Skin color (white; mulatto/black) Clinical symptoms (asymptomatic/mild; moderate/severe) Type of HAART (with PI; without PI)
0.95* 0.90* 0.84* 0.83* 0.81* 0.80* 0.78* 0.78* 0.76* 0.75*** 0.56* 0.29** 0.09 0.06
0.89* 0.85* 0.85* 0.81* 0.81* 0.75* 0.76* 0.75* 0.75* 0.75* 0.59* 0.36*** 0.14 0.08
Note: For categorical variables Spearman Rank Correlation was used. Abbr: aBMC, areal bone mineral content; BMD, bone mineral density; BMI, body mass index; HAART, combination antiretroviral therapy; HIV, human immunodeficiency virus; PI, protease inhibitors; UAMA, upper arm muscle area. *p ! 0.001; **p ! 0.05; ***p ! 0.01.
1 and 2 ( p O 0.05). Furthermore, Lin’s concordance correlation coefficients were of moderate to substantial magnitude. The linearity between the values measured by DXA and the estimates obtained with the models and also the analyses of
residuals are shown in Figs. 1 and 2, respectively. A random distribution of residuals for aBMD and BMC was observed for models 1 and 2, respectively (x2 5 3.05 and p 5 0.08; x2 5 1.60 and p 5 0.20).
Table 3 Multiple Regression Models for the Prediction of Bone Mass in Children and Adolescents Living With HIV (Florian opolis, Brazil, 2011) Multiple Regression Models Dependent variable Total aBMD (g cm2) Model 1* Y 5 0.1321904 þ (height 0.0028424) þ (age 0.0153856) þ (BMI 0.01837) þ (skin color 0.0412933) þ (sex 0.0273668) Dependent variable Total BMC (g) Model 2* Y 5 1193.955 þ (BW 45.70223) þ (age 32.50329) þ (AC 48.89363) þ (FD 12.30302) þ (sex 46.86491)
R2adjusted
0.84
0.94
SEE
0.048
109.28
RMSE
0.055
115.48
F
AIC n
BIC0
T
VIF
50.25
138.473
73.929
0.16
2.75
140.30
597.72
118.581
0.06
5.01
Note: Height in cm; age in years; skin color (1 5 white; 2 5 black/mulatto); sex (0 5 female, 1 5 male). Abbr: aBMD, areal bone mineral density; AC, arm circumference (cm); AIC, Akaike’s information criterion; BIC, Bayesian information criterion; BMC, bone mineral content; BMI, body mass index (kg m2); BW, body weight (kg); FD, femoral diameter (mm); HIV, human immunodeficiency virus; RMSE, root mean square error; SEE, standard error of the estimate; T, tolerance; VIF, variance inflation factor. *p ! 0.001. Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015
6
Lima et al. Table 4 Comparison Between Bone Mass Measured by DXA and Estimated With the Prediction Models (Models 1 and 2) in Children and Adolescents Living With HIV (Florianopolis, Brazil, 2011)
Measured and Estimated Bone Mass Total Total Total Total
aBMDDXA aBMDmodel 1 BMCDXA BMCmodel 2
Mean (SD) 0.886 0.886 1368.1 1368.1
(0.135) (0.125) (459.3) (446.1)
Rho ( p)
TE
AD
0.92 (!0.001)
0.007
!0.001
0.97 (!0.001)
!0.001
15.59
Note: There was no difference between the mean values obtained by DXA and those estimated with the models (t test 5 0.0000). Abbr: aBMD, areal bone mineral density; AD, average difference between measured and estimated values; BMC, bone mineral content; DXA, dual-energy X-ray absorptiometry; HIV, human immunodeficiency virus; Rho, Lin’s concordance correlation coefficient; SD, standard deviation; TE, total error.
Internal validation with 1000 bootstrap samples produced a corrected R2 of 84% and 94% to aBMD and BMC models, respectively.
Discussion The present study demonstrated that height, body weight, arm circumference, BMI, and femoral diameter, with age and skin color adjusted by sex, predict in 84% and 94% of aBMD and BMC, respectively, in children and adolescents living with HIV. These variables are accurate and reliable and can be easily measured in clinical setting. DXA, which is the preferred method for the evaluation of aBMD and BMC (23), was used as reference for the development of these models. These results are relevant to children and adolescents vertically infected with HIV living in Brazil, but with subsequent validation these models could be useful for bone mass monitoring elsewhere. Databases of healthy subjects can provide the parameters for the calculation of Z-scores (12) and consequently, the clinical interpretation of BMC and aBMD for age.
Previous studies have reported anthropometric variables as potential predictors of bone mass in children, (8,24,25) adolescents, (5,26) adults, and the elderly (7). As a previous research (24), we founded that height was the best predictor of aBMD in children and adolescents of both sexes. However, in a study with adolescent girls (26) fat-free mass presented the highest correlation with aBMD at different anatomical sites and was also the best predictor of aBMD in other study explaining 26% of bone mass (25). However, body weight was an important predictor of aBMD in children and young adults with cerebral palsy (8) and the correlations of body weight, height, BMI, and fat-free mass with aBMD were low to moderate in other study (25). We found that BMI, height, and age had a similar predictive power of aBMD, while body weight was the best predictor of BMC, followed by arm circumference, femoral diameter, and age. Anthropometric measurements, when combined, can show excellent predictive power of bone mass, and this was shown in prepubertal subjects (24) when height, diameters, skinfolds, and circumferences explained 60%e91% of aBMD and BMC. Our study corroborates these findings. The models explained
Fig. 1. Relationship between bone mass measured by DXA (aBMDDXA and BMCDXA) and predicted by the equations developed (aBMDmodel 1 and BMCmodel 2) in children and adolescents living with HIV (Florianopolis, Brazil, 2011). aBMD, areal bone mineral density; BMC, bone mineral content; DXA, dual-energy X-ray absorptiometry; HIV, human immunodeficiency virus. Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015
Prediction of Bone Mass in HIV + Youths
7
Fig. 2. Analysis of residual scores of areal bone mineral density (A) and bone mineral content (B) measured by DXA and estimated with the prediction models in children and adolescents living with HIV (Florianopolis, Brazil, 2011). aBMD, areal bone mineral density; BMC, bone mineral content; DXA, dual-energy X-ray absorptiometry; HIV, human immunodeficiency virus. from 84% to 94% of aBMD and BMC variance, respectively; although most participants were adolescents. This fact would imply more variability in bone mass, which is also affected by a delay in growth and maturity in HIV (27). In a Brazilian study, on the prediction of aBMD in adolescents (25), sex, fat-free mass, and abdominal muscle resistance explained approximately 40% of aBMD. In stratified analysis, abdominal resistance and maximum oxygen uptake and fat-free mass were the best predictors in boys and girls, respectively. Moreover, in North American adolescents, leg strength and power were predictors of aBMD and BMC (26). Both studies demonstrated the dynamic interaction between morphology and function and the possible prediction by physical fitness. However, several studies developing predictive or explanatory models of bone mass have considered many variables that represent size, such as diameters and height, in addition to others that can be proxies for gravitational or tension overload such as body weight, arm circumference, and UAMA (8,24e26). This is in agreement with the theoretical assumption and biological plausibility of the measurement of areal aBMD by DXA (5,28) and of osteogenic stimuli to induce bone formation (29), respectively. Body measures and the protein content of fat-free mass increase with age (30), a fact that would explain their importance in prediction models. Moreover, black or mulatto skin color has been shown to be associated with higher BMC compared with white skin color (12). Moreover, the models accounted for the sexual dimorphism in bone mass (31). We recommend the use of both models for the estimation of BMC and aBMD. The analyses of linearity highlight the high predictive power of the 2 models. Furthermore, correlation coefficients of moderate to substantial magnitude were obtained. The residuals showed a random distribution of the estimated values, while the average difference between measured and estimated values was virtually zero. The bootstrap procedure guaranteed an internal validity. Model 1 adequately estimated aBMD, from 100 to þ100 g cm2, in 97.9% of the children and adolescents studied. In a
restricted range, from 50 to þ50 g cm2, approximately 70% of the participants presented accurate estimates of aBMD. The same was observed for the BMC values with model 2 accurately estimating (200 to þ200 g) BMC in 93.7% of the participants. Within a restricted range, 100 to þ 100 g, 66.6% of the participants had valid estimates. However, the literature indicates the use of BMC due to the greater reproducibility and the absence of errors compared to areal density (23,28). Despite some limitations that suggest caution in interpreting the data obtained with the regression models, such as regional Brazilian sample as reference. The strengths of the present study include the high coefficient of determination (84%e94%) and the low standard errors of the estimate. Furthermore, to develop the regression models was considered the practicality of clinical application and the analysis of statistical parameters for robustness and reliability of the estimates. The models were developed for children and adolescents living with HIV using a heterogeneous sample in terms of use of protease inhibitors, clinical and immunological symptoms, and viral load. In conclusion, the present study developed prediction models of BMC and aBMD for children and adolescents living with HIV based on anthropometric and demographic variables. The models show high predictive power and agreed with the DXA method as the SEE are virtually zero. These prediction models can be useful to monitor bone mass in patients living with HIV, especially in scenarios in which resources are limited because anthropometry is easily available.
Acknowledgment The authors acknowledge the patients, investigators, and study personnel participating in this study. Also, we would like to thank Coordination of Improvement of Higher Education Personal (CAPES) for the awarded scholarships and Clınica Imagem for the assistance with DXA scans.
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015
8
Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jocd.2015.06.001.
References 1. Hazra R, Siberry GK, Mofenson LM. 2010 Growing up with HIV: children, adolescents, and young adults with perinatally acquired HIV infection. Annu Rev Med 61:169e185. 2. Arpadi SM, Shiau S, Arpadi CM, Yin MT. 2014 Bone health in HIV-infected children, adolescents and young adults: a systematic review. J AIDS Clin Res 5:2. 3. Rizzoli R, Bonjour JP. 1999 Determinants of peak bone mass and mechanisms of bone loss. Osteoporos Int 9(Suppl 2): S17eS23. 4. Specker BL, Schoenau E. 2005 Quantitative bone analysis in children: current methods and recommendations. J Pediatr 146:726e731. 5. Katzman DK, Bachrach LK, Carter DR, Marcus R. 1991 Clinical and anthropometric correlates of bone mineral acquisition in healthy adolescent girls. J Clin Endocrinol Metab 73:1332e1339. 6. Belissari A, Roche AF. 2005 Anthropometry and ultrasound. In: Human Body Composition. Heymsfield SB, Lohman TG, Wang Z, Going SB, eds. 2a ed. Champaign: Human Kinetics, 109e127. 7. Casini A, Mohamed EI, Gandin C, et al. 2003 Predicting bone mineral density of postmenopausal healthy and cirrhotic Italian women using anthropometric variables. Dig Liver Dis 35:881e887. 8. Henderson RC, Kairalla J, Abbas A, Stevenson RD. 2004 Predicting low bone density in children and young adults with quadriplegic cerebral palsy. Dev Med Child Neurol 46:416e419. 9. United Nations Development Programme. Institute for Applied Economic Research. Jo~ao Pinheiro Foundation. Atlas of Human Development in Brazil. Available at: http://www.atlasbrasil.org. br/2013/pt/. Accessed: November 20, 2014. 10. Giuliano IDCB, De Freitas SFT, De Souza M, Caramelli B. 2008 Subclinic atherosclerosis and cardiovascular risk factors in HIVinfected children: PERI study. Coron Artery Dis 19:167e172. 11. Leonard CM, Roza MA, Barr RD, Webber CE. 2009 Reproducibility of DXA measurements of bone mineral density and body composition in children. Pediatr Radiol 39:148e154. 12. Kelly TL, Wilson KE, Heymsfield SB. 2009 Dual energy X-ray absorptiometry body composition reference values from NHANES. PLoS One 4:e7038. 13. Pan H, Cole TJ. LMSgrowth, a Microsoft Excel add-in to access growth references based on the LMS method. Version 2.77. http:// www.healthforallchildren.co.uk/; 2012. Accessed: September 12, 2011. 14. Lima LRA, Silva RCR, Giuliano ICB, et al. 2013 Bone mass in children and adolescents infected with human immunodeficiency virus. J Pediatr 1:91e99. 15. Lohman TG, Roche AF, Martorell R. 1991 Anthropometric Standardization Reference Manual. Abridged ed. Champaign: Human Kinetics.
Lima et al. 16. Onis M, Onyango AW, Borghi E, et al. 2007 Development of a WHO growth reference for school-aged children and adolescents. Bull World Health Organ 85:660e667. 17. Frisancho AR. 1981 New norms of upper limb fat and muscle areas for assessment of nutritional status. Am J Clin Nutr 34: 2540e2545. 18. Greulich WW, Pyle SI. 1959 Radiographic atlas of skeletal development of the hand and wrist. Stanford, California: Stanford University Press. 19. Tanner JM. 1962 Growth at adolescence. 2nd ed. Oxford: Blackwell Scientific Publications. 20. Centers for Disease Control and Prevention. 1994 Revised classification system for human immunodeficiency virus infection in children less than 13 years of age; Official authorized addenda: human immunodeficiency virus infection codes and official guidelines for coding and reporting ICD-9-CM. MMWR 1994; 43(No. RR-12):[1-10]. 21. Centers for Disease Control and Prevention. 2008 Revised surveillance case definitions for HIV infection among adults, adolescents, and children aged !18 months and for HIV infection and AIDS among children aged 18 months to !13 yearsd United States. MMWR Recomm Rep 57(No. RR-10):1e12. 22. Haukoos JS, Newgard CD. 2007 Advanced statistics: missing data in clinical researchepart 1: an introduction and conceptual framework. Acad Emerg Med 14:662e668. 23. Crabtree NJ, Arabi A, Bachrach LK, et al. 2014 Dual-energy X-ray absorptiometry interpretation and reporting in children and adolescents: the revised 2013 ISCD Pediatric Official Positions. J Clin Densitom 17:225e242. 24. Miller JZ, Slemenda CW, Meaney FJ, et al. 1991 The relationship of bone mineral density and anthropometric variables in healthy male and female children. Bone Miner 14:137e152. 25. Fonseca RM, de Franca NM, Van Praagh E. 2008 Relationship between indicators of fitness and bone density in adolescent Brazilian children. Pediatr Exerc Sci 20:40e49. 26. Witzke KA, Snow CM. 1999 Lean body mass and leg power best predict bone mineral density in adolescent girls. Med Sci Sports Exerc 31:1558e1563. 27. de Martino M, Tovo PA, Galli L, et al. 2001 Puberty in perinatal HIV-1 infection: a multicentre longitudinal study of 212 children. AIDS 15:1527e1534. 28. Prentice A, Parsons TJ, Cole TJ. 1994 Uncritical use of bone mineral density in absorptiometry may lead to size-related artifacts in the identification of bone mineral determinants. Am J Clin Nutr 60:837e842. 29. Judex S, Carlson KJ. 2009 Is bone’s response to mechanical signals dominated by gravitational loading? Med Sci Sports Exerc 41:2037e2043. 30. Malina RM, Bouchard C, Bar-Or O. 2004 Growth, maturation, and physical activity. Champaign, Illinois: Human Kinetics. 31. Lu J, Shin Y, Yen MS, Sun SS. [In Press, available online 18 October 2014] Peak bone mass and patterns of change in total bone mineral density and bone mineral contents from childhood into young adulthood. J Clin Densitom.
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume
-,
2015