Predictors of growth and body composition in HIV-infected children beginning or changing antiretroviral therapy

Abstract

Objectives

To describe growth and body composition changes in HIV+ children after initiating or changing ART and correlate these with viral, immune and treatment parameters.

Methods

Ninety-seven prepubertal HIV+ children were observed over 48 weeks upon beginning or changing ART. Anthropometry and BIA were compared to NHANES to generate z-scores and HIV-exposed, uninfected children from WITS. Multivariate analysis evaluated associations between growth and body composition and disease parameters.

Results

All baseline lean and fat mass measures were below those of controls from NHANES. Weight, height and FFM index (FFM/height²) z-scores increased over time (p=0.004, 0.037 and 0.027) and waist/height ratio z-score decreased (p=0.045), but BMI and % body fat z-scores did not change. Measures did not increase more than in uninfected WITS controls. In multivariate analysis, baseline height, mid-thigh circumference and FFM z-scores related to CD4% (p=0.029, p=0.008 and 0.020) and change in FFM and FFM index z-scores to CD4% increase (p=0.010 and 0.011). Compared to WITS controls, baseline difference in height and mid-thigh muscle circumference were also associated with CD4%. Case-control differences in change in both subscapular skinfold (SSF) thickness and the SSF/triceps skinfold ratio were inversely associated with viral suppression. No measures related to ART class(es) at baseline or over time.

Conclusions

In these HIV+ children, beginning or changing ART was associated with improved growth and lean body mass, as indicated by FFM index. Height and LBM relate to CD4% at baseline and over time. Altered fat distribution and greater central adiposity are associated with detectable virus but not ART class(es) received.

Introduction

Poor growth is a common manifestation of HIV infection in children,^1-5 the pathophysiology of which remains poorly understood. The importance of growth is underscored by the finding that height growth velocity predicts survival, regardless of plasma viral load [HIV-1 RNA (VL)], age, and CD4⁺ cell count.⁶ The relationships between growth, VL, immune function and antiretroviral therapy (ART) remain unclear. Conflicting data exist from both pre- and post-HAART eras^6-13 about whether VL is associated with growth. Most, but not all,^11-15 reports of children on PI therapy note improved linear and ponderal growth. Some data suggest an association with VL that is not independent of immune function.¹⁰ It is still unclear whether improved growth sometimes seen with treatment is primarily a result of immune restoration, improved viral control or yet another mechanism.

HIV infection and/or ART may also alter body composition which may help differentiate starvation (preferential loss of fat resulting from inadequate energy intake) from cachexia [loss of lean body mass (LBM)], generally accepted to be cytokine mediated. Data are conflicting about preservation of lean body mass (LBM) in HIV-infected children.^2,16 Altered fat distribution in HIV-infected persons, particularly those on antiretroviral treatment, may also occur.¹⁷ In particular, increased central adiposity has been reported in both HIV-infected adults and children,^17,18 and is of concern because of the known association with cardiovascular morbidities.¹⁹ Although limited information is available examining associations and predictors of body composition and fat distribution in prepubertal HIV-infected children, exposure to protease inhibitors is frequently noted in association with lipodystrophy.^20-23 Data regarding association with disease measures such as viral load and CD4%, however, are conflicting.^20,21

The objectives of this study were to a) describe growth and body composition changes in HIV-infected children over 48 weeks after beginning or changing ART; b) compare these changes in HIV-infected children to those in both U.S. population based data and matched, HIV-exposed, uninfected children; c) correlate growth and body composition changes with ART class(es) and changes in VL and CD4+ cell count percentage (%). We hypothesized that there is a clinically significant inverse correlation between changes in LBM and VL and a direct correlation between changes in LBM and CD4+ cell count % in children beginning or changing ART. We further hypothesized that there would be a greater increase in central adiposity in children begun on therapy containing protease inhibitors (PI) compared to those begun on non-PI regimens.

Methods

Subjects and Design

Pediatric AIDS Clinical Trials Group Protocol 1010 (PACTG P1010) was a multi-site, prospective 48 week cohort study of HIV-infected, pre-pubertal children aged 1 month to <13 years who were beginning or changing ART. The ART criteria for inclusion were one of the following scenarios: a) beginning any ART if ART naïve, b) beginning protease inhibitor (PI) based ART if PI naïve, or c) changing ART for virologic failure to a regimen including ≥2 new drugs. Exclusion criteria have been previously described but briefly included pubertal development, concurrent acute illness or treatment within 180 days of entry with medications known to affect growth or body composition, e.g. steroids.²⁴ Ethics committee approval was obtained from each participating institution as was written informed consent from the parent or legal guardian and assent from the child when appropriate. Accrual began in June 2000 and continued until March 2004.

Visits were at study entry (within 72 hours prior to ART initiation or change) and at 12, 24, 36 and 48 wks thereafter. At each visit, the following evaluations were performed by trained staff: interim history and physical examination including Tanner staging; anthropometry [weight, height, circumferences (waist, hip and limb) and skinfold thicknesses (triceps, thigh and sub-scapular)]; single frequency tetrapolar bioelectrical impedance analysis [(BIA) 50 KHz, UniQuest-SEAC BIM4 instrument; UniQuest Limited, Brisbane, Australia] of total body impedance, resistance, reactance, and phase angle; plasma VL (HIV-1 RNA) and CD4+ T-lymphocyte count; and 3 day diet record (24 hour intake by recall if 3 day record not performed). Mid-arm and thigh muscle circumferences were calculated using standard equations and used as a measure of lean body mass (LBM). BIA measures were used to calculate total body water (TBW, L), fat free mass (FFM, kg), and fat mass (FM, kg) using equations previously validated in HIV-infected and uninfected children, TBW= 0.725 + 0.475H²/R + 0.140W; FFM= [3.474 + 0.459 H²/R +0.064W]/[0.769 – 0.009A – 0.016S]; and FM=W-FFM, where H is height (cm), R is resistance (ohms), W is weight (kg), A is age (years), and S is sex, 1 for males and 0 for females.²⁵ For children <8 years of age, the resistance index (H²/R) was utilized as a measure of total body water.²⁶ Percent body fat was calculated from BIA as [FM (kg)/weight (kg)] ×100, and FFM adjusted for height was calculated using the FFM index (FFM/height² ratio).²⁷

Laboratory Analyses

Laboratories with approved performance in the NIAID Division of AIDS Virology and Immunology Quality Assurance Programs conducted HIV-1 RNA and CD4⁺ cell measurements.

Sample Size

A sample size of 100 was calculated on the primary response variables of mid-arm muscle circumference (MAMC) and triceps skinfold thickness (TSF). Based on a pilot study, 100 subjects would allow detection of a change to within 0.5% for MAMC and 9.2% for TSF with 95% confidence. One hundred subjects would provide 99% power to detect a difference in MAMC change of 2.5% between viral responders and non-responders and 88% power to detect a 15% difference in change between the two groups.

Statistical Analyses

Two distinct analytical approaches were utilized to take account of sex-, race/ethnicity- and age-related differences in measures of growth and body composition in uninfected children: 1) sex/race/ethnicity/age-adjusted z-scores were calculated using data from a large, nationally representative cross-sectional sample of children [the National Health and Nutrition Examination Survey 1999-2002²⁸ (NHANES)] and 2) a case-control approach was used in which each child in this study was matched to one or more HIV-exposed, uninfected controls from another study which was socio-demographically similar, the Women and Infants Transmission Study ²⁹ (WITS), who were followed longitudinally.

For the first analytical approach using data from NHANES, growth and body composition z-scores at baseline were derived by selecting all available children in the NHANES database of the same sex, race-ethnicity and age (to within ±3 months) as a child in this study (the P1010 child). Then, for each growth and body composition measure, the z-score for the P1010 child was calculated as [(P1010 child’s measurement) – (measurement of values for matched NHANES children)]/[standard deviation of values for matched NHANES children]. This was repeated for measurements at weeks 24 and 48. Growth and body composition measures were log transformed before calculation of z-scores as this gave distributions of values which were more symmetric than untransformed values. The only anthropometric measures performed in our population that were not available in NHANES subjects were mid-thigh skinfold thickness and calculated mid-thigh muscle circumference. In addition, z-scores for BIA measures were only derived for children ≥ 8 years of age, as BIA was measured in NHANES beginning at this age. Across the growth and body composition measures, the mean (sd) number of NHANES children used in calculating a z-score for each P1010 child ranged from 34.5 (9.0) to 40.5 (12.9). A total of 6819 children from NHANES contributed data in calculating z-scores for anthropometric variables, including 2769 children aged ≥ 8 years for BIA variables. Weight, height and BMI of these children from NHANES , were compared to reference CDC growth curves to obtain mean percentiles of this control population versus that reference standard.

For each growth and body composition measure, the univariate association was evaluated between the baseline z-score and each of the following measures of baseline disease status: CD4%, log₁₀ HIV RNA level, CDC clinical classification, and prior ART exposure (with or without a PI in regimen). Multivariable regression analysis measured the association of baseline z-score (dependent variable) with standard HIV-related disease predictor variables of baseline CD4%, log₁₀ HIV-1 RNA level, CDC classification, and prior ART exposure adjusting also for log of the ratio of caloric intake to estimated caloric need at baseline (though the results changed minimally without adjustment).

Longitudinally, the associations of change in z-score from baseline to week 48 of follow-up and change in CD4% over the same period, viral load at week 48 (detectable versus undetectable HIV-1 RNA RT-PCR with sensitivity of 400 copies/ml) and ART class initially received during study follow-up (PI-containing, NNRTI-containing or both), adjusting for baseline z-score as well as baseline CD4%, log₁₀ HIV-1 RNA and CDC clinical classification. Regression analyses were also adjusted for mean caloric intake (log ratio of caloric intake to estimated caloric need) of P1010 participants over the study period to evaluate if associations noted were independent of diet (though the results changed minimally without adjustment). Fat, protein and caloric intake were analyzed using Nutritionist IV software (Hearst Corporation, San Bruno, CA).

For the second analytical approach using data from WITS, for each P1010 child (“case”), up to 3 matched “control” children from WITS were identified. Children were first matched on sex and race/ethnicity. In addition, as WITS followed children longitudinally, a control had to have a study visit at the same age (within ±3 months) as the case’s P1010 baseline visit. As WITS evaluated the Tanner Stage of females ≥ 7 years old and males ≥ 9 years old, WITS controls in these age ranges also had to be pre-pubertal at that visit. A total of 129 matched controls for 72 cases were identified (1 - 3 matched controls per case); 22 of 38 children > 8 years of age had no matches identified. WITS had very few children older than 8 years of age, limiting utility of this control population in our older subjects.

For each growth and body composition measure, to take account of the matching in the statistical analysis, a case-control difference at baseline was calculated by subtracting the mean of the measurements for the matched controls from the case’s measurement. Univariate and multivariable associations between these differences and the case’s baseline disease status (CD4%, log₁₀ HIV-1 RNA and CDC classification) and prior ART exposure were evaluated using the same methods as for the analysis of z-scores described above, except that the multivariable analyses also included sex, race/ethnicity and age as predictor variables. For each case and matched WITS control, the change from baseline in a measure over 48 weeks was calculated, and then a case-control difference in that change was obtained. Multivariable associations of these differences then proceeded as for the analysis of changes in z-scores except that the multivariable analyses also included sex, race/ethnicity and age as predictor variables; only results from multivariable analyses are presented because of the dependence of many associations on these demographic factors.

To evaluate how the 129 uninfected, control children from WITS compared with children in the general population, z-scores were also calculated using the NHANES data in the same way that z-scores were calculated for children in the P1010 study population.

Results

One hundred five patients were recruited to achieve the desired sample size of 100, as 5 patients were found to be ineligible after study entry, because of pubarche (n=3), disallowed medication (n=1), or withdrawal of consent prior to initial data collection (n=1). Three additional patients were excluded as the entry visit occurred subsequent to the change in ART, resulting in a final sample size for analyses of 97. Six patients withdrew from the study prior to the 48 wk visit. Demographic and clinical characteristics of the study population are shown in Table 1. Briefly, the mean (sd) age at entry was 5.88 (3.63) years with 54% female, 61% black, non-Hispanic, and 48% CDC Clinical Class A or N; mean CD4+ cell % was 24.8 (12.5) and mean HIV RNA was 4.55 (0.89) log₁₀ copies/ml corresponding to a geometric mean 35,338 copies/ml. Nearly one-third (29%) of subjects were ART naïve and an additional 24% were PI naïve at study entry. At both 24 and 48 weeks slightly greater than half of the children had VL<400 copies/ml. During the study, all children were on a nucleoside reverse transcriptase inhibitor and 19% received a non-nucleoside reverse transcriptase inhibitor (NNRTI) without a protease inhibitor (PI), 20% received both a NNRTI and a PI, and 57% received a PI without a NNRTI. One child changed from a PI- to a NNRTI-containing regimen and one from a NNRTI- to a PI-containing regimen in the first 7 days; these two children were classified according to the regimen received after 7 days. Two other children started on a PI regimen but changed later in follow-up to an NNRTI-containing regimen and were classified according to the initial regimen. No other changes of drug class were reported. Twenty-five children experienced pubarche during the 48 weeks on study, 20 of who were Tanner stage 2 at the 48 week visit. Dietary intake data were available for 82 children; mean total fat intake exceeded national recommendations in only 2 of these children (2%) and all but one child consumed protein in quantities equal to or greater than recommended for age and weight.

Table 1.

Demographic and Clinical Characteristics of PACTG 1010 Study Population

	Number of subjects	Percent of subjects*
Age≠
1 mo - <18 mos	9	9.3
18 mos - <3 yrs	15	15.5
3 yrs - <8 yrs	33	34.0
8 yrs- <13 yrs	40	41.2
Gender
Male	45	46.4
Female	52	53.6
Race-Ethnicity
White, Non-Hispanic	11	11.3
Black, Non-Hispanic	59	60.8
Hispanic	26	26.8
Other/Unknown	1	1.0
CDC Clinical Stage# ≠
A,N	47	48.4
B	32	33.0
C	18	18.6
Prior Therapy≠
ART Naïve	28	28.9
PI-Naïve**	23	23.7
PI-Exposed	46	47.4
ART Regimen on Study
NRTI	4	4.0
NRTI/NNRTI	18	18.6
NRTI/NNRTI/PI	19	19.6
NRTI/PI	55	56.7
NRTI/NNRTI/FI	1	1.0

	Entry	48 wks	Entry	48 Wks
	Number of subjects		Percent of subjects*
CD4%
0 - < 15	17	8	18.9	9.1
15 - <25	29	14	32.2	15.9
≥25	44	66	48.9	75.0
Missing	7	9
HIV-1 RNA Copies/ml
< 400	2	46	2.2	54.1
400 - < 10,000	22	12	23.7	14.1
10,000 - < 30,000	15	11	16.1	12.9
30,000 - <100,000	26	10	28.0	11.8
≥ 100,000	28	6	30.1	7.1
Missing	4	12

^*Percent is of subjects with known status; Percents may not add to 100 due to rounding

^**ART exposed;

^≠At study entry; NRTI = nucleoside reverse transcriptase inhibitor; NNRTI = non-nucleoside reverse transcriptase inhibitor; PI = protease inhibitor; FI = Fusion Inhibitor (enfuvirtide);

All anthropometric measures and calculated TBW, FFM and percent body fat z-scores were significantly (p<0.05) below zero in HIV-infected children at baseline (study entry) as shown in Figure 1. Similarly, in comparison to the matched HIV-exposed, uninfected children from WITS, most measures were also significantly lower at entry, with the exception of mid-arm muscle circumference (MAMC), mid-thigh skinfold (MTSF) and % body fat which approached the limit of significance (0.05<p<0.1; Figure 2). When compared to NHANES data, the uninfected control children from WITS also had z-scores which were significantly lower than zero on multiple measures of fat at both baseline and 48 weeks, including triceps skinfold (TSF), subscapular skinfold (SSF) and body mass index (BMI), as well as for weight and waist-circumference at 48 weeks (data not shown). Mean (95% CI) weight, height and BMI percentiles of the NHANES controls on the CDC reference curve were 62.8 (61.0,64.5), 56.9 (55.2,58.5) and 65.2 (63.2,67.0) respectively, each greater than the reference population, p<0.001.

Figure 1.

Difference in Anthropometric and Body Composition Z-scores in P1010 Study Population Compared to Adjusted Norms from NHANES at Study Entry and Week 48

Figure 2.

Anthropometric and Body Composition Differences in P1010 Study Population Compared to Matched HIV-exposed, Uninfected WITS Controls at Study Entry and Week 48

Over the 48-week course of therapy, mean (sd) weight [0.16 (0.53), p=0.004], height [0.14 (0.61), p=0.037], FFM [0.27 (0.48), p=0.001] and FFM index [FFM/height², 0.30 (0.81), p=0.027] z-scores increased significantly (Figure 1) while the waist/height ratio z-score decreased [− 0.19 (0.79), p=0.045]. At the 24-wk visit, there was a significant increase in mean z-scores for MAMC [0.28 (1.22), p=0.033] and mid-thigh circumference [MTC, 0.16 (0.45), p=0.030]. The latter changes, however, were no longer significant at the 48-week visit. By contrast, there was no significant difference in change at 48 weeks between cases and matched HIV-exposed, uninfected controls from WITS (Figure 2).

In multivariate analyses of baseline z-scores (NHANES controls), more severe stunting was associated with CDC Clinical Class B and C compared to N or A (height z-score −0.56, p=0.044 and −1.06, p=0.002 respectively) and a higher waist/height ratio z-score with Class C, p=0.006 (See Table 2). Baseline z-score for height, mid-thigh circumference (MTC) and FFM were each associated with baseline CD4% (z-scores 0.19, 0.38 and 0.38 higher per 10% greater CD4%, p=0.029, 0.008 and 0.020 respectively), as shown in Table 2. Fat-free mass index, however, was not associated with CD4%, p=0.22. Viral load at baseline was significantly associated only with lower peripheral fat stores as measured by mean TSF z-score of −0.19 per each log₁₀ RNA copies/ml higher, p=0.043. Similarly, in multivariate analysis of the differences at entry between P1010 cases and WITS controls (Table 3), case-control difference in height and mid-thigh muscle circumference (MTMC) were both associated with baseline CD4%, (compared with uninfected HIV-exposed matched controls, mean height and MTMC in infected children were higher by 1.6 cm and 1.32 cm, respectively, per 10% higher baseline CD4% of the infected child, p=0.015 and 0.019 respectively). In addition, compared with uninfected HIV-exposed matched controls, mean BMI was higher by 3.03 kg/m² in infected children with CDC Category C disease compared with those with CDC Category A/N disease (p=0.029). In the comparison with WITS controls, there were no significant associations at baseline between any growth or body composition measure and VL. Nor were significant associations seen with ART or PI exposure, although the difference in triceps skinfold thickness in PI-exposed vs. ART naïve children approached significance (− 4.54 mm, p=0.057).

Table 2.

Association of Baseline Growth and Body Composition Z-scores with Measures of Disease Status in P1010 Study Population

Baseline value	CD4% (per 10% higher)	HIV RNA (per 1 log10 copies/ml higher)	CDC category (vs. category A/N)
			Category B	Category C
Weight (n=96)	0.04 (−0.11 , 0.19 ) p= 0.56	0.03 (−0.13 , 0.19 ) p= 0.74	−0.31 (−0.80 , 0.17 ) p= 0.20	−0.44 (−1.03 , 0.15 ) p= 0.14
Height (n=96)	0.19 (0.02 , 0.36 ) p= 0.029	0.10 (−0.08 , 0.28 ) p= 0.29	−0.56 (−1.10 , −0.02 ) p= 0.044	−1.06 (−1.72 , −0.39 ) p= 0.002
Mid-arm Muscle Circumference (n=95)	0.25 (−0.03 , 0.53 ) p= 0.075	0.02 (−0.26 , 0.30 ) p= 0.89	0.41 (−0.47 , 1.30 ) p= 0.35	0.54 (−0.55 , 1.62 ) p= 0.33
Mid-thigh circumference (n=41)	0.38 (0.10 , 0.65 ) p= 0.008	−0.08 (−0.33 , 0.16 ) p= 0.49	0.16 (−0.58 , 0.91 ) p= 0.66	0.14 (−0.80 , 1.08 ) p= 0.76
Triceps Skinfold (n=95)	−0.08 (−0.25 , 0.09 ) p= 0.36	−0.19 (−0.38 , −0.01 ) p= 0.043	−0.12 (−0.68 , 0.44 p= 0.67	−0.36 (−1.04 , 0.32 ) p= 0.29
Subscapular Skinfold (n=96)	0.04 (−0.12 , 0.21 ) p= 0.60	−0.12 (−0.30 , 0.05 ) p= 0.17	0.06 (−0.47 , 0.59 ) p= 0.82	0.05 (−0.60 , 0.70 ) p= 0.88
Subscapular/Triceps Skinfold Ratio (n=95)	0.29 (−0.02, 0.60) p= 0.069	0.09 (−0.24, 0.42) p= 0.59	0.21 (−0.80, 1.23) p= 0.68	−0.02 (−1.25, 1.21) p= 0.98
Fat Free Mass (n=41)	0.38 (0.06 , 0.70 ) p= 0.020	0.09 (−0.19 , 0.37 ) p= 0.51	−0.39 (−1.25 , 0.48 ) p= 0.37	−0.37 (−1.46 , 0.72 p= 0.49
FFM/Height2 Ratio (n=41)	0.19 (−0.12, 0.49) p= 0.22	−0.03 (−0.30, 0.24) p= 0.80	−0.19 (−1.02, 0.63) p= 0.64	0.20 (−0.84, 1.25) p= 0.70
Waist/height ratio (n=82)	0.03 (−0.14 , 0.19 ) p= 0.76	−0.11 (−0.29 , 0.07 ) p= 0.23	0.31 (−0.23 , 0.86 ) p= 0.26	0.90 (0.27 , 1.53 ) p= 0.006
Body Mass Index (n=96)	−0.08 (−0.23 , 0.07 ) p= 0.28	−0.02 (−0.17 , 0.14 ) p= 0.82	−0.18 (−0.65 , 0.29 ) p= 0.46	0.06 (−0.51 , 0.64 ) p= 0.83
Body fat % (n=41)	0.22 (−0.02 , 0.45 ) p= 0.073	−0.02 (−0.23 , 0.19 ) p= 0.86	0.14 (−0.51 , 0.79 ) p= 0.66	0.22 (−0.60 , 1.04 ) p= 0.59

Bolded results are significant (p<0.05); n is the number of observations included in the analyses; Models were adjusted for log of the ratio (caloric intake/estimated caloric need) at baseline; prior ART exposure was also included in the model but was not significant for any growth or body composition measure; Z-scores derived from matched children from NHANES.

Table 3.

Association of Baseline Case-Control Differences and Measures of Disease Status

Baseline differences (case-control)	CD4% (per 10% higher)	HIV RNA (per 1 log10 copies/ml higher)	CDC category (vs. category A/N)
			Category B	Category C
Weight kg (n=64)	1.10 (−0.43, 2.63) p= 0.16	0.35 (−2.18, 2.88) p= 0.78	−0.02 (−5.37, 5.33) p= 0.99	3.26 (−2.66, 9.19) p= 0.27
Height cm (n=64)	1.60 (0.33, 2.88) p= 0.015	0.05 (−2.06, 2.17) p= 0.96	−2.67 (−7.15, 1.80) p= 0.24	−3.25 (−8.23, 1.73) p= 0.20
Mid-arm muscle circumference cm (n=64)	0.07 (−0.46, 0.61) p= 0.78	−0.09 (−0.97, 0.78) p= 0.83	−0.63 (−2.47, 1.21) p= 0.49	1.51 (−0.56, 3.58) p= 0.15
Mid-thigh muscle circumference cm (n=57)	1.32 (0.23, 2.41) p= 0.019	−0.05 (−2.00, 1.89) p= 0.96	1.04 (−2.88, 4.96) p= 0.60	1.93 (−2.47, 6.32) p= 0.38
Mid-thigh skinfold mm (n=63)	−0.71 (−2.72, 1.30) p= 0.48	−1.18 (−4.94, 2.58) p= 0.53	2.15 (−5.05, 9.35) p= 0.55	5.18 (−2.88, 13.24) p= 0.20
Triceps skinfold mm (n=66)	−0.14 (−1.29, 1.01) p= 0.81	0.03 (−1.93, 1.20) p= 0.97	0.41 (−3.66, 4.48) p= 0.84	3.20 (−1.38, 7.78) p= 0.17
Subscapular skinfold* mm (n=64)	0.25 (−0.83, 1.34) p= 0.64	0.07 (−1.81, 1.94) p= 0.94	−0.03 (−3.85, 3.79) p= 0.99	2.98 (−1.31, 7.28) p= 0.17
BMI kg/m2 (n=63)	−0.13 (−0.82, 0.57) p= 0.72	0.34 (−0.81, 1.49) p= 0.56	1.08 (−1.37, 3.52) p= 0.038	3.03 (0.32, 5.74) p= 0.029
Body fat % (n=14)	−1.11 (−24.36, 22.15) p= 0.86	−0.33 (−21.59, 20.92) p= 0.95	−1.13 (−81.64, 79.39) p= 0.96	1.03 (−100.59, 102.65) p= 0.97
Resistance Index (H2/R) cm2/ohms (n=47)	1.02 (−0.83, 2.88) p= 0.27	−0.62 (−3.30, 2.05) p= 0.64	−4.47 (−10.09, 1.14) p= 0.12	0.02 (−6.06, 6.09) p= 1.00
Waist-height ratio (n=49)	−0.02 (−0.03, 0.00) p= 0.054	−0.01 (−0.02, 0.01) p= 0.53	0.03 (−0.03, 0.08) p= 0.38	0.04 (−0.02, 0.11) p= 0.15

Bolded results are significant (p<0.05); Case-control differences derived from matched children in WITS; Results are from multivariate models which also adjusted for sex, race/ethnicity, baseline age and baseline log calorie (intake/estimated need); prior ART exposure was also included in the model but was not significant for any growth or body composition measure; n is the number of observations included in the analyses.

Multivariate analysis of the z-score changes at 48 weeks revealed an association between change in FFM as well as the FFM index (FFM/height² ratio) and CD4%; for each 10% increase in CD4% over the 48 weeks there was an associated mean (95% CI) increase in FFM z-score of 0.42 (0.11, 0.73), p= 0.010 and FFM index z-score of 0.57 (0.14, 1.00), p= 0.011. As with baseline measures, there were no differences in adjusted z-score changes with PI vs. NNRTI vs. PI and NNRTI based HAART regimens on study.

Similar multivariate analysis of the difference in change between cases and matched WITS control children revealed greater change in case-control difference in truncal fat as measured by sub-scapular skinfold thickness (SSF) and truncal/limb fat ratio (subscapular/triceps skinfold ratio) for children whose viral load was detectable at 48 weeks (4.07 mm, p=0.001 and 0.12 mm, p=0.036 respectively).

When results were not adjusted for caloric intake all described statistically significant associations based on z-scores or on case-control differences remained statistically significant.

Discussion

Our hypothesis that increases in LBM would be directly associated with improved CD4% was proven by the increase in FFM index of 0.57 z-score for each 10% increase in CD4% at 48 weeks. The associations between case-control difference in MTMC and CD4% at entry in the WITS comparison and the MTC z-score and CD4% at entry in the NHANES comparison, lends further credence to this association. There was, however, no evidence to support our hypothesis that viral suppression would relate to improvements in LBM. We did, however, find an association between higher persistent viral load and fat distribution. Greater increase in truncal fat (measured by SSF) and trunk/limb fat ratio (SSF/TSF) relative to controls in the WITS comparison was seen in children who did not achieve viral suppression compared to those who did. Higher VL at baseline has been shown to predict loss of both extremity and truncal fat in HIV-infected adults, ³⁰ the loss of extremity fat with higher viral burden is similar to the finding we noted between smaller TSF and higher viral load at entry.

It is unclear how improved CD4% cell counts might relate physiologically to improved muscle mass. An association between an increase in extremity muscle mass and increase in CD4+ cell count has been previously reported in adults.³⁰ One could speculate that lower CD4% may be related to intercurrent infections, and subsequent loss of LBM from catabolism due to these infections. Authors of the previous study speculated that it may reflect ‘improved health, nutrition and mobility’ resulting from improved CD4+ cell count. Improved nutrition seems an unlikely explanation given the finding persists when adjusting for caloric intake in our study, but again, reducing intercurrent infections could reduce nutritional needs.

The children in this study had similar mean gains in height z-score (0.14), but greater gains in weight z-score (0.16), compared to those previously described for children on PI therapy.¹² These improvements occurred in the first 48 weeks on therapy and were independent of viral suppression, contrary to a previous report that improved growth was delayed until 96 weeks on therapy, and only for virologic responders.¹¹ Height increases appear greater than that seen with PI therapy in a study by Miller et al.,¹⁵ although only adjusted z-scores are presented; our populations differ in that the P1010 children were receiving a variety of different HAART regimens, which may have resulted in greater overall effect. Growth and body composition changes in our study were independent of class(es) of ART begun at study entry. Additionally, there was no evidence that there was an increase in central adiposity in the study population as a whole as reflected by mean waist/height ratio z-score, which actually decreased over the 48 weeks, or by SSF. Nor was there evidence to support our hypothesis that PI therapy would be associated with a greater increase in central adiposity.

Our findings on body composition at baseline do not concur with those of Fontana et al.¹⁶ in that % body fat z-score was significantly lower than the comparison children in NHANES at entry, and there was a trend towards the same in comparison to the HIV-exposed children in WITS [mean (sd) z-score = −0.51 (0.69) and case-control difference vs. WITS −5.6 % (11.5), p<0.001 and p=0.09 respectively] suggesting that FM was more diminished in these children than was lean mass. This suggests there may be a component of relative ‘starvation’ in addition to the impaired anabolism demonstrated by lower measures of LBM. Alternatively, it could be that the NHANES controls had greater relative body fat compared to Fontana’s controls. The latter possibility is supported by the mean BMI percentile of matched NHANES controls used in this study of 65.2%.

In our study population, both FFM and FFM index z-scores increased significantly, suggesting greater lean mass in the population as a whole was not entirely a result of greater linear growth but rather there was also a relative increase in muscle mass. Percent body fat and BMI did not change, however; apparently a corresponding appropriate gain in fat mass also occurred. Unfortunately, the significant increase of arm muscle circumference seen in our population at 24 wks was not sustained. Nor was there greater gain in arm or thigh muscle circumference (or any anthropometric or BIA measure) in our population when compared to control children from WITS, despite entering the study with smaller measures of both muscle and fat stores. Apparently the anabolic response that may result in improved linear growth does not result in significantly greater muscle circumference in the children as a group, at least over 48 weeks. Miller reported that PI treatment was independently associated with improvements in LBM as measured by arm muscle circumference,¹⁵ a finding we did not corroborate either looking at differences between PI exposed vs. naïve children at baseline or comparing those receiving PI to non-PI based HAART regimens during the study. Results to date in WITS also demonstrate a trend towards decreased arm and thigh muscle masses in infected vs. uninfected children, with no evidence that this is changing in the era of HAART.³¹

There are several limitations to this study. It is likely the HIV-infected children in our study differed from the overall US population represented in NHANES data in ways for which we could not adjust; differences between the WITS uninfected children and the NHANES population in several anthropometric measures support this speculation. Furthermore, BIA measures were only available in children >8 years of age in NHANES, limiting utility of BIA in this comparison. NHANES itself is cross-sectional data and not ideal for comparison with data from subjects followed longitudinally. The HIV-exposed, uninfected cohort in WITS is likely to be more similar to our study population than the overall population in NHANES, but the case-control method did not allow generation of z-scores; there were also few matches for the older children. Results of the two comparisons are discrepant in some cases; it is likely that some of these differences are due to the different ages represented, as age was significantly associated with multiple measures at both baseline and over the 48 weeks. Other differences may be the result of fewer available matched children in the WITS cohort, resulting in less power to detect changes in case-control differences over time that may be clinically significant. The subjects in our study also began diverse ART regimens, limiting the power to detect changes that may be associated with specific ART class(es). Although we did not find an association with specific ART classes, all children were on treatment, so it is not possible to sort out the contribution that treatment per se may have to growth and body composition changes. The lack of associations at entry with PI therapy compared to ART or PI naivety suggests that there may not be substantive effects of ART per se on growth or body composition. There are also many comparisons such that some findings of borderline significance may have occurred by chance. Finally, we did not have a comparison group of HIV-infected children that were not beginning or changing therapy, so clearly the associations noted may be different in children on long-term therapy. The major strengths of the study include a sample size larger than most prospective studies evaluating body composition changes in this population, a homogenously pre-pubertal population at study entry, evaluation before and after ART initiation or change, and collection of both anthropometric and BIA measures.

In summary, in this population of HIV-infected children predominantly with mild to moderate disease, initiation or change in ART was followed by improvements in linear and ponderal growth as well as improved FFM index, when compared to population-based norms, but not compared to matched HIV-exposed, uninfected children. These differences in results according to comparison group may primarily be a difference related to age, as younger children were disproportionally represented in the comparison to exposed, uninfected children, or of power, as there were fewer matched children in the latter group. Limb muscle mass circumferences did not improve significantly nor were there changes in lean:fat ratios over time in the group as a whole. Height and other measures of lean body mass were associated with CD4% at study entry and over time and greater truncal fat is associated with failure to achieve viral suppression. Further investigation is required to understand the physiologic relationships underlying these associations.

List of Abbreviations

ART

antiretroviral therapy

HIV

human immunodeficiency virus

NHANES

National Health and Nutrition Examination Survey

WITS

Women and Infants Transmission Study

FFM

fat free mass

LBM

lean body mass

BIA

bio-electrical impedance analysis

MAMC

mid-arm muscle circumference

MTMC

mid-thigh muscle circumference

BMI

body mass index

PI

protease inhibitor

NNRTI

non-nucleoside reverse transcriptase inhibitor

VL

viral load

HAART

highly active antiretroviral therapy

TBW

total body water

FM

fat mass

R

resistance

H

height

W

weight

A

age

S

sex

NIAID

National Institute for Allergy and Infectious Diseases

AIDS

acquired immunodeficiency syndrome

RT-PCR

reverse transcriptase polymerase chain reaction

MTSF

mid-thigh skinfold

TSF

triceps skinfold

SSF

subscapular skinfold

MTC

mid-thigh circumference