Full text of DOI:10.1086/315865

1385
Treatment-Mediated Changes in Human Immunodeficiency Virus (HIV) Type 1
RNA and CD4 Cell Counts as Predictors of Weight Growth Failure, Cognitive
Decline, and Survival in HIV-Infected Children
Jane C. Lindsey,1 Michael D. Hughes,2
Ross E. McKinney,4 Mary K. Cowles,5
Janet A. Englund,6,7,a Carol J. Baker,6,7
Sandra K. Burchett,3 Mark W. Kline,7 Andrea Kovacs,8
and Jack Moye9
1Center for Biostatistics in AIDS Research and Department
2
3
of Biostatistics, Harvard School of Public Health, and Children’s
Hospital, Harvard Medical School, Boston, Massachusetts; Pediatric
AIDS Clinical Trials Unit, Duke University Medical Center, Durham,
4
5
North Carolina; Department of Statistics and Actuarial Science,
6
University of Iowa, Iowa City; Departments of Microbiology
and Immunology and Pediatrics, Baylor College of Medicine,
Houston, Texas; Maternal Child Program, University of Southern
California Medical Center, Los Angeles; National Institute of Child
7
8
9
Health and Human Development, National Institutes of Health,
Rockville, Maryland
This meta-analysis of 5 large studies of the Pediatric AIDS Clinical Trials Group was
undertaken to evaluate the predictive value of antiretroviral treatment-mediated changes in
3 markers of human immunodeficiency virus (HIV) type 1 disease progression—HIV-1 RNA
level, CD4 cell count, and CD4 percentage—for weight growth failure, cognitive decline, and
survival in HIV-infected children. Proportional hazards models were used to assess the prog-
nostic value of the markers at baseline and after 24 weeks of treatment, with data from 1089
children. Among children receiving nucleoside with or without nonnucleoside reverse-tran-
scriptase inhibitors, higher immunologic and lower virologic markers at baseline and after 24
weeks were significant independent predictors of survival, whereas virologic markers were
significant predictors of weight growth and cognitive failure in children 11 year old. The
finding of differential age effects on pediatric-specific clinical outcomes emphasizes the need
for continued investigation of treatment effects in children.
The value of human immunodeficiency virus (HIV) type 1
markers of immunologic function [5], virus load [6], and rates
of disease progression [7] are associated with age. This often
makes it inappropriate to extrapolate findings in adult studies
to children. Cohort studies in HIV-infected infants and children
[6, 8] have reported the predictive value of HIV-1 RNA levels,
CD4 cell counts, and CD4 percentages on disease progression
(defined as incidence of a new Centers for Disease Control and
Prevention category C diagnosis [9] or death). However, there
has been limited study of the association between treatment-
mediated changes in these parameters and disease progression
in children. One study reported the prognostic value of these
predictors at baseline and after 24 weeks of therapy on a com-
posite clinical outcome (weight growth failure, neurologic/
neuropsychologic decline, development of significant new HIV-
related diagnoses, and death [10, 11]). Mofenson et al. [12]
described the effect of treatment-mediated changes in CD4 per-
centage and HIV-1 RNA level on survival.
We report the results of a meta-analysis of studies of the
Pediatric AIDS Clinical Trials Group (PACTG), undertaken to
evaluate the predictive value of antiretroviral treatment-medi-
ated changes in HIV-1 RNA levels, CD4 cell counts, and CD4
percentages on weight growth failure, cognitive decline, and
survival in children. Although the drug combinations used in
these studies did not include protease inhibitors, nucleoside
reverse-transcriptase inhibitors (NRTIs) are the backbone of
RNA levels and CD4 cell counts as prognostic markers for
progression to AIDS or death is clearly established in HIV-
infected adults in natural history studies [1, 2] and clinical trials
[3, 4]. Unlike adults, who are developmentally stable and in
whom changes in immunologic status are predominantly caused
by the HIV infection, children experience changes in growth
and in central nervous system (CNS) and immune function.
Disease progression manifests itself differently in children, and
Received 24 March 2000; revised 17 July 2000; electronically published
22 September 2000.
Presented in part: 7th Conference on Retroviruses and Opportunistic In-
fections, San Francisco, February 2000 (abstract N32e).
Informed consent was obtained from children or their parents and guard-
ians, and human experimentation guidelines of the US Department of
Health and Human Services were followed in the conduct of this clinical
research.
Financial support: Statistical and Data Management Center of the Pe-
diatric AIDS Clinical Trials Group (National Institute of Allergy and In-
fectious Diseases cooperative agreement AI-41110).
a
Present affiliation: Department of Pediatrics, University of Chicago,
Chicago, Illinois.
Reprints or correspondence: Dr. J. C. Lindsey, Center for Biostatistics in
AIDS Research, Harvard School of Public Health, 651 Huntington Ave.,
Boston, MA 02115 (lindsey@sdac.harvard.edu).
The Journal of Infectious Diseases 2000;182:1385–93
᭧ 2000 by the Infectious Diseases Society of America. All rights reserved.
0022-1899/2000/18205-0013$02.00

1386
Lindsey et al.
JID 2000;182 (November)
Bayley or McCarthy tests. Weight growth failure and cognitive
decline are part of the composite clinical end points used in PACTG
trials and have similar definitions to those used in ongoing trials.
Statistical methods. Analysis of variance, adjusting for study,
was used to compare baseline, week 24, and changes from baseline
to week 24 marker levels, by age group [24]. To assess the prognostic
value of the markers, proportional hazards models [25] were used
to evaluate associations between marker levels and time from the
week 24 HIV-1 RNA measurement to each clinical end point. HIV-
1 RNA levels and CD4 cell counts were analyzed as continuous
variables on the base 10 logarithmic (log10) scale because this gave
good satisfaction of the assumptions required for the proportional
hazards models. All models were stratified by treatment and study.
This allows the hazard for the outcome to be different for each
study and treatment, but assumes that the association between
outcome and differences in marker levels at baseline or week 24 is
the same across studies. Initial exploratory models were run to see
whether sex, race/ethnicity, or prior treatment influenced the as-
sociations of the markers with the risk of any of the clinical out-
comes. None were found. Results were illustrated by categorizing
HIV-1 RNA levels and CD4 cell counts. Tests for interaction were
used to evaluate whether any associations varied between the dif-
ferent treatments used in the studies. Comparisons of differences
in likelihoods from nested proportional hazards models were used
to assess the relative strength of the association between the 2 CD4
measures and the clinical end points. Follow-up for survival was
censored at the time of study closure or at loss to follow-up. Follow-
up for weight growth failure and cognitive decline was censored at
the time of the last weight or neuropsychologic test score mea-
surement, respectively.
current highly active antiretroviral combinations and may be
useful in countries with no or limited access to protease inhib-
itors. Our results provide important information concerning the
association between treatment-mediated virologic and immu-
nologic marker changes and disease progression in children,
how these associations vary with age, and how they might differ
from results in HIV-1–infected adults. Our findings may be
useful in designing future clinical trials and for the management
of HIV-infected infants and children.
Methods
Studies and evaluations. Studies eligible for inclusion in the
meta-analysis (PACTG 152 [7], PACTG 239 [13], PACTG 240 [14],
PACTG 245 [15], and PACTG 300 [16]) were randomized clinical
trials of NRTIs and the nonnucleoside RTI (NNRTI) drug nevi-
rapine, conducted by the PACTG. All studies had у24 weeks of
follow-up and HIV-1 RNA measurements in at least some children
before the start of therapy and after 24 weeks of study. Lymphocyte
subsets were measured at baseline and at 24 weeks. Weights were
measured every 4 weeks, and cognitive test scores were recorded
at age-appropriate intervals (every 24 weeks in children !30 months
old and at least every year in older children).
The focus of the analysis was on baseline and week 24 marker
values. The baseline value was the marker measurement closest to
and within 30 days before the start of study treatment. Because of
the variation in clinic visits in different studies, the week 24 mea-
surement was defined as the value closest to week 24 within a
window of time between 16 and 32 weeks after the start of treat-
ment. Only children with HIV-1 RNA levels and CD4 cell counts
at both baseline and week 24 were included in the analysis. Children
also had to have started у1 new agent to which they had never
been previously exposed. If a subject participated in 11 of the 5
studies, data from the first enrollment were used. Althoughdifferent
assays for quantifying HIV-1 RNA levels were used, the same assay
was used for all measurements within each study; thus, stratifying
the analyses by study controlled for this source of variation. Values
below the limit of quantification of an assay were replaced by the
limit. Only 3% of values were below the limit, so this practice had
minimal effect on the results.
Outcome measurements. Because the definition of disease pro-
gression varied between studies, 3 end points with objective defi-
nitions were used. The first was death occurring 124 weeks after
the initiation of study antiretroviral therapy. Weight growth failure,
which could only occur 132 weeks after initiation of treatment,
was defined as the third of 3 consecutive 6-month weight growth
velocities less than the third percentile for age and sex [17], with
weight at the third time point less than the 10th percentile for age
and sex. Cognitive decline, which could occur 124 weeks after
baseline, was defined as a decline in scaled score 130 points within
the Bayley I [18] or II [19] tests; a decline in scaled score 130 points
from a Bayley I or II to any other test; a decline in scaled score
115 points from any previous McCarthy [20], Wechsler preschool
and primary scale of intelligence [21], Wechsler intelligence scale
for children, third edition [22], or Wechsler intelligence scale for
children, revised [23]; or no increase in raw score from the previ-
ously administered test of the same type for 112 weeks for the
Results
Patient population. Five studies were included in the anal-
ysis. Children started treatment between August 1991 (PACTG
152) and June 1997 (PACTG 239 and 300); follow-up was com-
pleted between August 1995 (PACTG 152) and September 1997
(PACTG 239). A total of 2081 children started treatment on
у1 study, but only 1089 are included in this analysis: 150 did
not receive a new antiretroviral, 166 had !24 weeks of follow-
up, and 676 had incomplete baseline or week 24 virologic or
immunologic data. Three hundred eleven of the 676 subjects
were enrolled in older studies, in which RNA testing was done
only on a subset of subjects. There were no statistically signif-
icant differences between the study population and those not
included in the analysis in the distributions of sex, race/ethnic-
ity, age, CD4 cell count, or source of infection. There was a
slightly higher proportion of treatment-naive subjects in the
analysis subset (73% vs. 67%), which was a function of the
eligibility criteria requiring у1 new antiretroviral treatment.
Table 1 shows the treatments evaluated in the 5 studies and
baseline characteristics of the 1089 infants and children. All
studies used either monotherapy (47% of patients) or combi-
nation therapy with NRTIs. Two treatment arms of PACTG
245 included the NNRTI nevirapine in combination with 1 or
2 NRTIs. Among the 1089 children, 50% were boys and 55%

JID 2000;182 (November)
HIV Disease Progression in Children
1387
Table 1. Baseline characteristics of 1089 participants in 5 Pediatric AIDS Clinical Trial Group (PACTG) clinical trials.
PACTG studya
152
239
240
245
300
Total
Characteristic
(n p 413)
(n p 36)
(n p 77)
(n p 178)
(n p 385)
(n p 1089)
Treatment
Monotherapy
Zidovudine
d4T
129 (31)
31 (40)
46 (60)
160 (15)
46 (4)
ddI
Combination
Zidovudine ϩ ddI
Zidovudine ϩ 3TC
ddI ϩ NVP
144 (35)
140 (34)
16 (44)
139 (36)
299 (28)
20 (56)
34 (19)
84 (22)
162 (42)
278 (26)
162 (15)
69 (6)
69 (39)
75 (42)
Zidovudine ϩ ddI ϩ NVP
75 (7)
Age
р6 Months
16 Months–1 year
11–2 Years
12–5 Years
15 Years
33 (8)
82 (20)
69 (17)
102 (25)
127 (31)
36 (100)
10 (13)
25 (33)
12 (16)
21 (27)
9 (12)
41 (11)
57 (15)
62 (16)
104 (27)
121 (31)
120 (11)
168 (15)
152 (14)
263 (24)
386 (35)
4 (2)
9 (5)
36 (20)
129 (73)
Race/ethnicity
White, non-Hispanic
Black, non-Hispanic
Hispanic/Latino
Other
60 (15)
224 (54)
119 (29)
10 (2)
6 (17)
22 (61)
5 (14)
3 (8)
15 (20)
48 (62)
14 (18)
0 (0)
40 (23)
68 (38)
70 (39)
0 (0)
60 (16)
239 (62)
85 (22)
1 (0)
181 (17)
601 (55)
293 (27)
14 (1)
Sex
Male
Female
214 (52)
199 (48)
18 (50)
18 (50)
44 (57)
33 (43)
103 (58)
75 (42)
160 (42)
225 (58)
539 (50)
550 (50)
Source of infection
Perinatal
Not perinatal
Unknown
372 (90)
40 (10)
1 (0)
36 (100)
76 (99)
1 (1)
345 (90)
18 (5)
22 (6)
829 (76)
59 (5)
23 (2)
Not collected
Previous use of antiretrovirals
Yes
178 (100)
178 (16)
24 (6)
389 (94)
17 (47)
19 (53)
15 (20)
62 (81)
173 (97)
5 (3)
71 (18)
314 (82)
300 (28)
789 (73)
No
CD4 cell count,b cells/mm3
CD4 percentageb
HIV-1 RNA,b log10 copies/mL
Weight-for-age z-scoreb
Cognitive scaled score, %
!50
790 (453, 1281)
24 (16, 31)
5.2 (4.6, 5.8)
Ϫ0.7 (Ϫ1.6, 0.1)
1914 (1242, 2550)
37 (28, 43)
5.7 (5.3, 6.2)
Ϫ1.3 (Ϫ2.0, Ϫ0.4)
974 (630, 1478)
25 (18, 32)
4.4 (3.8, 4.8)
Ϫ0.8 (Ϫ1.8, 0.0)
65 (17, 194)
6 (2,11)
5.0 (4.5, 5.3)
Ϫ1.2 (Ϫ2.0, Ϫ0.0)
685 (360, 1175)
23 (14, 31)
5.1 (4.5, 5.6)
Ϫ0.4 (Ϫ1.4, 0.4)
653 (238, 1194)
22 (11, 30,
5.1 (4.5, 5.6)
Ϫ0.7 (Ϫ1.6, 0.2)
9
15
76
7
13
80
8
2
72
12
12
76
7
17
75
9
16
76
50–69
у70
NOTE. Data are no. (%), except where noted. 3TC, lamivudine; d4T, stavudine; ddI, didanosine; HIV, human immunodeficiency virus; NVP, nevirapine.
a
Percentages represent proportion of children included in analysis database from each study.
Data are median (lower quartile, upper quartile).
b
were black (non-Hispanic). The studies targeted different age
groups with varying histories of prior antiretroviral use. Over-
all, 73% of the children had no previous treatment. In the 4
studies that collected data on source of infection, 90%–100%
of the children were perinatally infected. There was a wide range
of HIV-1 RNA levels and immunologic values before initiation
of study treatment across studies. In all studies, the median
weight-for-age z-score was !0, (i.e., !50th percentile of weight
for age and sex in children without HIV infection). In all, 25%
of the children recorded cognitive scaled scores !70 on the age-
appropriate measures of cognitive functioning: 9% were !50.
Median follow-up from the week 24 HIV-1 RNA measure-
ment for survival varied from a median of 0.7 years in PACTG
239 to 2.2 years in PACTG 152 (overall median, 1.2 years; table
2). Follow-up for the weight growth and cognitive end points
was slightly shorter, with overall medians of 1.0 and 0.9 years,
respectively, reflecting that evaluations were only obtained at
defined clinic visits and were less likely to occur after discon-
tinuation of study treatment prior to the end of study. All
follow-up was used to maximize the number of events.
The number of end points occurring on each study is shown
in table 2. Some children were not included in the weight growth
failure analyses because they had to stay in the study for у32
weeks to satisfy the end point definition. Similarly, since cog-
nitive decline was defined as a decline in scores, only children
with у2 valid tests could meet the end point. Overall, weight
growth failure and cognitive decline occurred more frequently
(12% and 13%) than death (5%). The rates showed variability
over studies, reflecting, in part, the different disease status and
ages of the study populations.

1388
Lindsey et al.
JID 2000;182 (November)
Table 2. Clinical end point rates in Pediatric AIDS Clinical Trials Group (PACTG) studies.
PACTG study
End point
152
239
240
245
300
All
55/1089 (5) 1.2
123/1031 (12) 1.0
123/938 (13) 0.9
Death
Weight growth failure
Cognitive decline
35/413 (9) 2.2
80/406 (20) 1.9
69/366 (19) 1.8
2/36 (6) 0.7
2/28 (7) 0.6
6/25 (24) 0.6
3/77 (4) 0.9
7/74 (10) 0.8
6/58 (10) 0.8
11/178 (6) 0.7
13/151 (9) 0.5
13/134 (10) 0.5
4/385 (1) 0.9
21/372 (6) 0.8
29/355 (8) 0.8
NOTE. Data are no. of children with clinical end points for those who could have reached an end point after 24 weeks of
treatment, percentage (in parentheses), and median follow-up from week 24 human immunodeficiency virus–1 RNA measurement,
in years.
Associations of markers and disease progression with age.
2). Among infants р1 year old at baseline, the estimated Kap-
lan-Meier percentages for death within 1 and 2 years after week
24 were 6% and 12%. This compares with 1% and 7% among
children 11 year old.
Median marker values at baseline, after 24 weeks, and changes
from baseline to week 24 are summarized in table 3 for each
of 5 age groups. Similar to previous reports, HIV-1 RNA levels
and CD4 cell counts and percentages were significantly lower
in older children (P ! .01) at baseline and at week 24 (P ! .01).
There were no statistically significant associations between the
changes in any of these markers from baseline to week 24 and
age.
Eighteen-month event rates (estimated from Kaplan-Meier
curves) for the 3 clinical end points are also shown in table 3.
There were significant differences in the rates of death, weight
growth failure, and cognitive decline end points across ages
(P ! .01), with older children tending to have lower rates. Dif-
ferences between rates among age groups 11 year old tended
to be small, compared with the difference between children 11
year and р1 year old. Therefore, results concerning the eval-
uation of the prognostic value of markers focused primarily on
whether any of the associations found varied between these 2
age groups.
In age-adjusted analyses, there was a significant association
between the risk of death and higher HIV-1 RNA level at week
24 (P ! .001) and between the risk of death and lower log10
CD4 cell counts (P ! .001) and CD4 percentages (P ! .001).
There was no evidence that either CD4 measure was more
highly associated with the risk of death. These associations
between the risk of death and HIV-1 RNA level and CD4 cell
counts (categorized into 2 groups) are illustrated in table 4 (rows
and columns labeled “Total”) and were apparent in children
р1 and 11 year old. The association of risk of death and cat-
egorized HIV-1 RNA level is shown graphically in figure 1. The
graphs portray clear differentiation with HIV-1 RNA level in
both age groups.
In multivariate proportional hazards models that included
age, HIV-1 RNA, and 1 of the 2 CD4 measures at week 24
(modeled as continuous variables), the associations with each
marker were both significant (P ! .01). The association of HIV-
1 RNA level and CD4 cell count by age group is illustrated in
table 4. Children with both higher HIV-1 RNA level (у5 log10
Prognostic value of the markers for survival. Of the 1089
infants and children, 55 (5%) died during a median follow-up
of 1.2 years after their week 24 HIV-1 RNA measurement (table
Table 3. Median marker levels, median changes from baseline to week 24, and event rates 18 months after
week 24, by age.
Age at start of treatment
р6 Months
(n p 121)
16 Months–1 year
(n p 167)
11–2 Years
(n p 153)
12–5 Years
(n p 263)
15 Years
(n p 385)
Marker
Pa
HIV-1 RNA, log10 copies/mm3
Baseline
Week 24
Baseline to week 24
CD4 cell count, cells/mm3
Baseline
5.88
5.40
0.40
5.69
5.20
0.52
5.26
4.77
0.40
4.96
4.36
0.43
4.72
4.38
0.28
!.01
!.01
.09
1482
1339
Ϫ130
1247
1247
4
1073
987
Ϫ69
726
681
14
265
293
6
!.01
!.01
.24
Week 24
Baseline to week 24
CD4 percentage
Baseline
29
32
0
26
28
3
24
28
3
23
27
3
13
16
1
!.01
!.01
.27
Week 24
Baseline to week 24
18 Month event rate
Survival
Weight growth failure
Cognitive decline
6
28
41
7
18
34
5
12
16
4
14
10
5
13
10
!.01
!.01
!.01
NOTE. HIV, human immunodeficiency virus.
a
P for age effects adjusted for study: analysis of variance for markers and log rank test for clinical end points.

JID 2000;182 (November)
HIV Disease Progression in Children
1389
Table 4. Clinical end point rates by 18 months after human im-
munodeficiency virus (HIV)–1 RNA measurement.
associations of the risk of death and these markers. As with
death, there were significantly increased risks for weight growth
failure with higher HIV-1 RNA levels and lower levels of each
CD4 measure at week 24 (P ! .001, univariate analyses adjusted
for age). However, although the association with HIV-1 RNA
was clear among children 11 year old, it was not apparent
among younger infants (P ! .001 for difference in association
between age groups). This is illustrated in figure 2, which shows
a clear separation between event rates defined by level of HIV-
1 RNA at week 24 in children 11 year old at baseline, but no
separation among those р1 year old. For CD4 cell count, the
association was not significantly different between the 2 age
groups. These trends are also seen in table 4. There is a much
smaller difference in 18-month event rates between the 2 cat-
egories of HIV-1 RNA level in the р1-year-old age group than
in the older group, but there is a clear separation in the event
rates between the categories of CD4 cell count in both age
groups. In addition, on the basis of differences in the log like-
lihood between models with log10 CD4 cell count and CD4
percentage relative to those with only 1 immunologic param-
eter, there was strong evidence that the log10 CD4 cell count
was more highly associated with the risk of weight growth fail-
ure than was CD4 percentage.
In multivariate proportional hazards models, these results
persisted: there was a significant difference between age groups
in the association between HIV-1 RNA level at week 24 and
the risk for weight growth failure, with no apparent effect in
infants р1 year old. Lower CD4 cell count at week 24 was still
significantly associated with an increased risk of weight growth
failure independently of HIV-1 RNA level. Week 24 CD4 per-
centage was no longer prognostic with HIV-1 RNA already in
the model. There was no evidence of differences in the asso-
ciation of the immunologic end points and outcome by age
group. In contrast to the results for death, adding either baseline
HIV-1 RNA level or CD4 cell count into a model that already
had the week 24 levels of these markers did not add significantly
to the model. Therefore, marker values observed at week 24
and not the baseline values are most important in determining
the risk of weight growth failure. However, after taking into
account the week 24 HIV-1 RNA levels and CD4 cell counts,
the risk of weight growth failure was 2.1 times greater in infants
р1 year old than in older infants and children 11 year old
End point, age
!500 CD4 у500 CD4
group, and HIV-1 RNA cells/mm3,
cells/mm3,
KM %
Total,
No.
No.
log10 copies/mL
KM %
KM % events children
Death
р1 Year old
!5
у5
Total
0
25
22
3
5
4
3
10
7
3
21
24
120
168
288
11 Year old
!5
у5
Total
3
23
10
1
0
1
1
14
4
8
23
31
595
206
801
Weight growth failure
р1 Year old
!5
у5
Total
40
48
50
18
17
17
19
23
22
15
31
46
114
154
268
11 Year old
!5
у5
Total
13
46
24
4
16
6
7
32
13
34
43
77
574
189
763
Cognitive decline
р1 Year old
!5
у5
Total
11 Year old
!5
0
45
41
34
37
35
33
39
36
22
43
65
107
137
244
5
26
12
9
19
11
8
23
12
28
30
58
520
174
694
у5
Total
NOTE. Rates are Kaplan-Meier (KM) estimates of percentage of children
having clinical end points, by age group at entry and HIV-1 RNA and CD4 cell
counts at week 24.
copies/mL) and lower CD4 cell count (!500 cells/mm3) had the
highest risk of death. Event rates were lower among children
with HIV-1 RNA level !5 log10 copies/mL and/or у500 CD4
cells/mm3.
When we included baseline HIV-1 RNA levels and CD4 mea-
sures in the multivariate model, these were also significant or
marginally significant predictors of the risk of death in addition
to values at week 24 (table 5). Thus, it is not only the level of
these markers after 24 weeks of treatment that is important for
assessing progression but also their baseline marker values.
Age was not a significant predictor of the hazard of death
in the multivariate model (table 5). Furthermore, the inclusion
of age by marker interaction terms showed there was no sig-
nificant evidence that the associations of any of the markers
with the hazard of death varied by age. Hence, the differences
in mortality between the 2 age groups are largely explained by
the differences in HIV-1 RNA and CD4 cell levels at baseline
and week 24. Thus, the risk ratio for death comparing 2 children
whose HIV-1 RNA levels differ by 1 log10 copies/mL or whose
CD4 cell counts differ by 1 log10 cells/mm3 does not depend on
whether the 2 children are !1 or 11 year old.
( ). Thus, the difference in risk of weight growth failure
P p .01
between the younger and older children is not fully explained
by differences in marker levels.
Prognostic value of the markers for cognitive decline. Over-
all, cognitive decline had the highest event rates (table 2). Chil-
dren р1 year old had an 18-month event rate of 36% versus
12% in children 11 year old at baseline (table 4). In univariate
models that controlled for age, only week 24 HIV-1 RNA level
was strongly predictive of cognitive decline (P ! .001). There
was marginally significant evidence that the association was
weaker in infants р1 year old than in older children (P p
Prognostic value of the markers for weight growth failure.
The associations between the risk for weight growth failure and
HIV-1 RNA and CD4 measures were quite different from the

1390
Lindsey et al.
JID 2000;182 (November)
Figure 1. Cumulative proportion of deaths, by week 24 human immunodeficiency virus (HIV)–1 RNA level, among 1089 subjects participating
in 5 Pediatric AIDS Clinical Trials Group trials.
.06 for difference in association between 2 age groups). This
can be seen in table 4, where the difference in event rates in
the р1-year-old age group between those with week 24 HIV-1
RNA levels !5 versus у5 log10 copies/mL (33% vs. 39%) is
smaller than the difference in the older age group (8% vs. 23%).
Cumulative proportions of cognitive decline according to cat-
egory of week 24 HIV-1 RNA level showed more separation
in children 11 year old, compared with the younger children
(figure 3). CD4 measures at week 24 were only marginally sig-
nificant (log10 CD4 cell count, P p .09; CD4 percentage,
P p .11).
In the multivariate model that included age, HIV-1 RNA
level, and each of the 2 CD4 measures, week 24 CD4 measures
were not significant. Furthermore, neither HIV-1 RNA levels
nor the CD4 measures at baseline were significant when added
to a model that included the week 24 values. Age remained a
significant predictor with week 24 HIV-1 RNA levels in the
model; thus, as with weight growth failure, the increased event
rates in the younger children are not fully explained by the
higher HIV-1 RNA levels.
power to assess the prognostic value of the marker levels across
types of clinical outcomes and across age groups. Conclusions
did not vary by treatment or study, although there was little
power to address this question directly. For survival, both HIV-
1 RNA and CD4 levels after 24 weeks of antiretroviral therapy
were independently predictive, which is consistent with previous
findings combining clinical end points or death [10] and for
survival alone [12]. Our results also show that these marker
values before initiation of antiretroviral treatment provide ad-
ditional predictive power. This confirms the findings by Pal-
umbo et al. [11] for PACTG 152 and suggests that clinicians
need to consider both baseline and during-treatment HIV-1
RNA level and CD4 marker levels when making patient care
decisions. For example, more aggressive therapies in children
with poorer baseline marker levels could be considered. In adult
patients, there have been conflicting reports regarding the im-
portance of baseline HIV-1 RNA values after adjusting for
levels achieved after initiation of treatment [2, 3, 26]. This dif-
ference between children and adults may reflect the fact that,
in children, marker values change with age, even in the absence
of treatment [6].
Another important finding regarding survival in infants and
children is the lack of a significant age effect once HIV-1 RNA
and CD4 levels (CD4 cell count or percentage) at week 24 and
baseline are considered. Thus, the higher mortality rates in
infants р1 year old, compared with older children, appear to
be explained by differences in marker values with age. This
finding has implications for clinical practice, because it suggests
that the difference in risk of death approximately doubles with
each 1 log10 copy/mL increase in HIV-1 RNA level at week 24
and at baseline (table 5). This observation was also found when
Discussion
This study demonstrates the independent value of HIV-1
RNA and CD4 measures obtained 24 weeks after initiation of
new antiretroviral treatment for predicting survival of HIV-
infected children. It also shows that associations between these
surrogate markers and the risk of weight growth failure or
cognitive decline differs from their association with the risk of
death and may differ by age.
By combining data across 5 pediatric trials, we gained greater

JID 2000;182 (November)
HIV Disease Progression in Children
1391
Table 5. Risk ratios (95% confidence intervals) from multivariate proportional hazards models describing the association between human
immunodeficiency virus (HIV)–1 RNA, immunologic function (log10 CD4 cell count or CD4 percentage), and death.
CD4 measure in model
Agea
Week 24 RNAb
Baseline RNAb
Week 24 CD4
Baseline CD4
Log10 CD4c
CD4%d
1.30 (0.53–3.17), P p .57 2.26 (1.31–3.90), P p .003 2.11 (1.17–3.80), P p .01 2.06 (1.10–3.87) P p .03 1.78 (0.93–3.41 ) P p .09
1.16 (0.50–2.67), P p .73 2.32 (1.34–4.01), P p .003 1.92 (1.08–3.42), P p .03 1.20 (0.99–1.46), P p .06 1.21 (0.98–1.50), P p .08
a
Relative risk (RR) associated with age р1 year relative to age 11 year.
RR associated with each 1 log10 higher HIV-1 RNA level.
RR associated with each 1 log10 lower CD4 cell count.
RR associated with each 5% lower CD4 percentage.
b
c
d
differences in rates of a composite clinical end point across age
groups were analyzed [10].
reliable after children discontinued treatment, making it less
likely that a weight growth failure or cognitive decline end point
would be identified. However, altering the weight growth failure
definition to require only 2 or 1 weight growth velocity mea-
surements below the third percentile did not influence the mag-
nitude of the differences between age groups. Also, using a
combined weight growth failure, cognitive decline, or death end
point did not eliminate the difference. Relying only on HIV-1
RNA levels and CD4 cell counts as end points in clinical trials
in young infants may not be optimal, because they do not seem
to capture important effects on weight growth and cognitive
failure. In children 11 year old, marker levels had similar effects
on weight growth failure, cognitive decline, and survival, and,
although it would be important to control for age in an analysis,
marker values at week 24 could be interpreted in an age-
independent manner.
The association between weight growth failure and cognitive
decline and HIV-1 RNA and CD4 levels in HIV-infected chil-
dren has not been well studied. In our analysis, because of the
large sample size, we were able to evaluate associations with
weight growth failure and cognitive decline as separate clinical
outcomes. For both, we found associations between higher
HIV-1 RNA levels at week 24 and the risk of weight growth
failure and cognitive decline, and these associations were par-
ticularly apparent in children 11 year old (figures 2 and 3).
However, among infants р1 year old, there appeared to be no
association between HIV-1 RNA levels and weight growth fail-
ure and only a weak association with risk of cognitive decline.
We did additional analyses to assess the impact of the dif-
ferential data collection schedules after children discontinued
the antiretroviral study treatment and the potential informative
censoring of the clinical end points by death and whether either
might cause the different associations between risk and HIV-1
RNA levels between age groups. Weight was recorded every 12
weeks, and collection of weight and cognitive tests became less
The prognostic value of the CD4 measures depended on the
clinical end point. When used in combination with HIV-1 RNA
level, both were equally prognostic for survival, but only the
CD4 cell count was predictive of weight growth failure, and
neither was predictive of cognitive decline, suggesting that the
Figure 2. Cumulative proportion of weight growth failures, by week 24 human immunodeficiency virus (HIV)–1 RNA level, among 1089
subjects participating in 5 Pediatric AIDS Clinical Trials Group trials.

1392
Lindsey et al.
JID 2000;182 (November)
Figure 3. Cumulative proportion of cognitive declines, by week 24 human immunodeficiency virus (HIV)–1 RNA level, among 1089 subjects
participating in 5 Pediatric AIDS Clinical Trials Group trials.
CD4 cell count is the most useful immunologic marker when
evaluating treatment efficacy. The fact that neither CD4 mea-
sure was predictive of cognitive decline may reflect a direct effect
of HIV-1 on the CNS, rather than being a consequence of
immunosuppression. This would support the use of therapeutic
agents that cross into the CNS, especially when HIV-1 RNA
levels exceed 5.5 log10 copies/mL (∼30,000 copies/mL), because
the risk of cognitive decline was markedly increased above this
level (figure 3). This might be particularly important in infants
р1 year old, because we also found that younger age was pre-
dictive of the greater absolute risk of cognitive decline even
after adjustment for HIV-1 RNA level.
of life [27]. This emphasizes the need for continued investigation
of treatment effects on all aspects of disease progression in the
pediatric population.
Acknowledgments
We thank K. O’Donnell (Duke University Medical Center, Durham,
NC) and P. Brouwers (Texas Children’s Cancer Center, Baylor College
of Medicine, Houston) for helping to define the cognitive end points,
Shirley Traite (Center for Biostatistics in AIDS Research, Harvard
School of Public Health, Boston) for creating the analysis data sets,
the children and families who participated in the Pediatric AIDS Clin-
ical Trials Group clinical trials, and the reviewers for their helpful
comments.
Although these 5 studies included children from 3 months
to 18 years old, with a wide range of HIV-1 associated clinical
findings, they were conducted when the standard of care was
changing rapidly. Antiretroviral treatment-mediated changes in
HIV-1 RNA and CD4 cell levels were much smaller than those
observed during the protease inhibitor era, so it is not possible
to extrapolate our conclusions to children who are currently
enrolled in clinical trials. Our analysis population included a
subset of subjects with a slightly higher proportion of treatment-
naive subjects who had tolerated treatment and had not ex-
perienced disease progression for у24 weeks. However, with
these caveats, it is clear that age is an important factor and
that conclusions about associations of surrogate markers with
clinical outcomes in adult studies cannot be universally applied
to the pediatric population, especially to infants р1 year old.
The differential impact of these markers on the 3 clinical end
points of death, weight growth failure, and cognitive decline is
also important because treatments aimed at altering marker
levels may improve prognosis for one outcome but have less
impact on another aspect of the disease directly affectingquality
References
1. Mellors JW, Munoz A, Giorgi JV, et al. Plasma viral load and CD4ϩ lym-
phocytes as prognostic markers in HIV-1 infection. Ann Intern Med
1997;126:929–38.
2. Riddler SA, Mellors JW. HIV-1 viral load and clinical outcome: review of
recent studies. AIDS 1997;11:S141–8.
3. Marschner IC, Collier AC, Coombs RW, et al. Use of changes in plasma
levels of human immunodeficiency virus type 1 RNA to assess the clinical
benefit of antiretroviral therapy. J Infect Dis 1998;177:40–7.
4. O’Brien WA, Hartigan PM, Martin D, et al. Changes in plasma HIV-1 RNA
and CD4ϩ lymphocyte counts and the risk of progression to AIDS. N
Engl J Med 1996;334:426–1673.
5. Denny T, Yogev R, Gelman R, et al. Lymphocyte subsets in healthy children
during the first 5 years of life. JAMA 1992;267:1484–8.
6. Shearer WT, Quinn TC, La Russa P, et al. Viral load and disease progression
in infants infected with HIV-1. N Engl J Med 1997;336:1337–42.
7. Englund JA, Baker CJ, Raskino C, et al. Zidovudine, didanosine, or both
as the initial treatment for symptomatic HIV-infected children. N Engl J
Med 1997;336:1704–12.
8. Abrams JA, Weedon J, Steketee RW, et al. Association of human immu-

JID 2000;182 (November)
HIV Disease Progression in Children
1393
nodeficiency virus (HIV) load early in life with disease progression among
bined zidovudine-lamivudine versus didanosine monotherapy in children
with symptomatic, therapy-naive, HIV-1 infection. J Pediatr 1998;133:
500–8.
HIV-infected infants. J Infect Dis 1998;178:101–8.
9. Centers for Disease Control and Prevention. 1994 revised classificationsystem
for human immunodeficiency virus infection in children less than 13 years
of age; official authorized addenda: human immunodeficiency virus in-
fection codes and official guidelines for coding and reporting ICD-9-CM.
MMWR Morb Mortal Wkly Rep 1994;43(RR-12):1–10.
17. Roche AF, Himes JH. Incremental growth charts. Am J Clin Nutr 1980;33:
2041–52.
18. Bayley N. Bayley scales of infant development. San Antonio, TX: Psycho-
logical, 1969.
10. Palumbo PE, Raskino C, Fiscus S, et al. Predictive value of quantitative
plasma HIV RNA and CD4ϩ lymphocyte count in HIV-infected infants
and children. JAMA 1998;279:756–61.
11. Palumbo PE, Raskino C, Fiscus S, et al. Virologic and immunologic response
to nucleoside reverse transcriptase inhibitor therapy among HIV-infected
infants and children. J Infect Dis 1999;179:576–83.
19. Bayley N. Bayley scales of infant development. 2d ed. San Antonio, TX:
Psychological, 1993.
20. McCarthy D. McCarthy scales of children’s ability. Cleveland: Psychological,
1970.
21. Wechsler D. Wechsler preschool and primary scale of intelligence: revised.
San Antonio, TX: Psychological, 1989.
12. Mofenson LM, Korelitz J, Meyer WA, et al. The relationship between serum
human immunodeficiency virus type 1 (HIV-1) RNA level, CD4 lympho-
cyte percent, and long-term mortality risk in HIV-1–infected children. J
Infect Dis 1997;175:1029–38.
22. Wechsler D. Wechsler intelligence scale for children. 3d ed. San Antonio, TX:
Psychological, 1991.
23. Wechsler D. Wechsler intelligence scale for children: revised. San Antonio,
TX: Psychological, 1974.
13. Kovacs A, PACTG Study Team 239. A phase I evaluation of the safety and
toxicity of ZDV and ddI in combination in HIV-infected or exposed in-
fants and a phase II study of the effect of ddI alone versus ZDV/ddI
combo on CD4 decline and growth in infants with HIV infection. Spon-
sored by the National Institute of Allergy and Infectious Diseases,PACTG
protocol version 5.0, April 1997.
14. Kline MW, VanDyke RB, Lindsey JC, et al. A randomized comparative trial
of zidovudine (ZDV, AZT) versus stavudine (d4T) in children with human
immunodeficiency virus infection. Pediatrics 1998;101:214–20.
15. Burchett S, PACTG Study Team 245. A comparative study of combination
antiretroviral therapy in children and adolescents with advanced HIV
disease. Sponsored by the National Institute of Allergy and Infectious
Diseases, PACTG protocol version 3.0, January 1996.
24. Rosner B. Fundamentals of biostatistics. 2d ed. Boston: Duxbury Press, 1986.
25. Cox DR. Regression models and life tables (with discussion). J R Stat Soc
Series B 1972;34:187–220.
26. Beacon HJ for the HIV Surrogate Marker Collaborative Group. Meta-anal-
ysis to investigate the value of HIV RNA level and CD4 cell count as
surrogate endpoints for evaluating treatments [abstract 0132]. In: Program
and abstracts of the 37th Interscience Conference on AntimicrobialAgents
and Chemotherapy (Toronto). Washington, DC: American Society for
Microbiology, 1997:132.
27. Nachman S, Lindsey JC, Dankner W, et al. Lack of improvement in growth
in HIV-infected children on HAART [abstract 0120]. In: Program and
abstracts of the 39th Interscience Conference on Antimicrobial Agents
and Chemotherapy (San Francisco). Washington, DC: American Society
for Microbiology, 1999:120.
16. McKinney R, Johnson GM, Stanley K, et al. A randomized study of com-