The Relationship between Serum Human Immunodeficiency Virus Type 1 (HIV-1) and Long-Term Mortality Risk in HIV-1Infected Children

In 20 -30 old men in Jakarta 2000

By: YOLANDA NABABAN 030.08.260

FACULTY OF MEDICINE TRISAKTI UNIVERSITY JAKARTA 2009

PREFACE

This paper titled The Relationship between Serum Human Immunodeficiency Virus Type 1 (HIV-1) RNA Level, CD4 Lymphocyte Percent, and Long-Term Mortality Risk in HIV-1-Infected Children in 20 -30 old men in Jakarta 2000 was created for the purpose of completing the assignment for the Medical English II in Trisakti University, Faculty of Medicine. In this paper is discussed all the information about the relationship between HIV-1 RNA level and CD4 lymphocyte percent, as well as their prognostic values, that may be affected by the pathogenesis of HIV infection. By reading the content of this paper, the reader will become more aware of the fatality of this disease and will take the steps to prevent these disease from happening. Many thanks and appreciation to those who have helped in the process of making this paper. Furthermore, my apologizes if there are errors contained in this paper for it is created during a learning process.

Jakarta, 12 04-2009

Yolanda Nababan

ii

TABLES OF CONTENTS Preface.......................................................................................................................................ii I. Introduction...................................................................................................................1 I.1 Background..............................................................................................................1 I.2 Methods....................................................................................................................2 I.3 Result........................................................................................................................4 I.4 Discussions.............................................................................................................13 I.5 References...............................................................................................................18

INTRODUCTION
iii

I.1 Background Primary human immunodeficiency virus type 1 (HIV-1) infection among adults is characterized by an initial burst of viremia, with HIV-1 RNA levels as high as 1O6-107 copies/mL, followed by a 100- to 1000-fold decline over the subsequent 2-3 months until reaching a steady-state plateau that may persist for years [1 -3]. A number of studies among infected adults have demonstrated that plasma RNA levels after seroconversion are independently predictive of risk for HIV-1 disease progression; higher HIV-1 RNA levels correlate with more rapid disease progression and elevated mortality risk, while treatmentinduced declines in RNA levels have been associated with clinical benefit [4-6]. These studies have led to the development of treatment recommendations for infected adults based on measurement of plasma HIV-1 RNA; RNA levels exceeding a threshold of ~ 10,000 copies/mL are viewed as indicative of the need to consider initiating therapy [7, 8]. Fewer data exist regarding HIV-1 RNA levels among infected children, particularly regarding the relationship of RNA with long-term clinical outcome. Several studies of perinatally infected children have shown persistent high RNA levels (s= 106 copies/mL) throughout the first 2 years of life, with only 2- to 10-fold declines from initial peak values during the first 12 months of life; these high levels may be observed despite normal CD4 lymphocyte counts and lack of symptoms [9-13]. RNA levels then appear to fall slowly until 24-36 months of age, independent of antiretroviral treatment and immunologic or clinical status [10-12]. These findings may reflect the influence of an immature but developing immune system that requires several years before achieving the capacity to control viral replication to an extent similar to that observed in infected adults. In perinatally infected children, RNA levels that are extremely elevated after 1-2 months of age (>300,000 copies/ mL) have been associated with more rapid progression to AIDS early in life. However, because of significant overlap in RNA levels between rapid and non-rapid progressors, no single threshold level predictive of disease progression has yet been identified in children [11, 12]. The National Institute of Child Health and Human Development (NICHD) Intravenous Immunoglobulin (IVIG) Clinical Trial was initiated to evaluate the effect of IVIG versus albumin placebo for prophylaxis of bacterial infections in HIV-1-infected
iv

children. Although mortality was similar between study groups, reduction in the incidence of infections and hospitalizations, as well as slowing of decline in CD4 lymphocytes, was observed with IVIG treatment [14-17]. During the course of this trial, blood was collected from patients at baseline and every 3 months during the study and sera were stored in a central repository. Additionally, follow-up data on the vital status of enrolled children 5 years after completion of the trial were obtained. Thus, these stored sera provide a unique opportunity to retrospectively evaluate HTV-1 KNA levels in a large HIV-1-infected pediatric cohort with well-defined long-term clinical outcomes.

The objectives of the current analysis were to describe the distribution of baseline HIV-1 RNA levels at study entry and changes over time in this cohort of children with mild to moderate HIV-1 disease, to evaluate the association of long-term mortality risk with baseline RNA level and subsequent RNA levels over time, and to evaluate the independent contributions of HIV-1 RNA level and CD4 lymphocyte percent to long-term mortality risk.

I.2 Methods The NICHD IVIG Clinical Trial was a randomized, double-blind placebocontrolled phase III outpatient clinical trial of the use of IVIG for prophylaxis of bacterial infections. The trial was conducted between March 1988 and January 1991 in 28 clinical centers in the mainland United States and Puerto Rico. Three hundred seventy-six HIV-1 -infected, nonhemophiliac children between the ages of 1 month and 12 years were enrolled, the majority of whom acquired infection perinatally and had only mild to moderate symptoms of HIV-1 disease. IVIG (Gamimune N; Miles, Berkeley, CA) was administered every 28 days to 187 patients (400 mg/kg of body weight), while 189 patients received a visually indistinguishable placebo consisting of 0.1% albumin without preservatives in 10% maltose, administered in an identical fashion. Children were seen monthly for examination and infusion. During these visits, information regarding intercurrent infections and medications was collected. Prophylaxis against Pneumocystis carinii pneumonia with a 3 consecutive dayper-week regimen of tri-methoprim-sulfamethoxazole and use of zidovudine antiretroviral therapy at any time after study entry were permitted, according to the prevailing medical standard of care as determined by the patient's physician with consent of the parent or
v

guardian. Further study methods and results are given elsewhere [14-17]. Additionally, in September 1996, each clinic site was contacted and asked to provide current vital status information for their study patients. Blood was collected from patients at entry to the trial and every 3 months thereafter during the course of the trial. Serum was separated, stored at 20 to 70C, and periodically shipped to a central repository on dry ice, where specimens were stored at 70C. All frozen specimens from children who had available baseline samples were retrieved and tested for HIV-1 RNA, using the nucleic acid sequence-based amplification (NASBA) assay according to manufacturer's instructions (Organon Teknika, Durham, NC). Specimens had been stored for 4-7 years before testing. All assays were performed by a single laboratory that was participating in the AIDS Clinical Trials Group HIV-1 Virology Quality Assurance program [18]. Testing was done over a 6-month period in late 1995 and early 1996. The NASBA assay involves extraction of nucleic acid by binding to silicon dioxide particles. The process can be used for testing many body fluids, including serum and plasma. HIV-1 RNA is amplified by an isothermal amplification procedure and quantified by coamplification with internal kit RNA calibrator standards of 10,000, 100,000, and 1,000,000 copies/mL. The quantity of amplified RNA was measured by means of an electrochemiluminescence technique, and results were expressed as copies of HIV-1 RNA per milliliter. The assay can detect up to a 4 log,0 variation in copy number; for the input volume of 100 yuL of serum used in this study, the lower limit of detection was 4000 copies/mL. Evaluation of interassay precision at the central laboratory indicated an interassay SD of 0.18 Iog10. CD4 lymphocyte percent and absolute count were measured at study entry and every 3 months during the trial. Flow cytometry was performed locally at laboratories participating in one of several national quality assurance programs. Procedures for performance of flow cytometric evaluations during the study have been described [16]. Because CD4 lymphocyte percent exhibits less measurement variability than the absolute count and varies less by age [19], CD4 cell percent rather than absolute count was used in the analyses.

The evaluation included baseline HIV-1 RNA copy number, baseline CD4 lymphocyte percent, and two measures of change in HIV-1 RNA copy number: absolute
vi

change (final value - baseline value) and yearly rate of change (slope of least squares regression line). The relationship between HIV-1 RNA level and age was assessed with a mixed-effects repeated-measures model. Intrapatient changes from baseline were assessed with the t test. The relationship of baseline HIV-1 RNA level and changes in RNA levels with mortality was evaluated with the %2 test for linear trend and Kaplan-Meier analyses. Proportional hazard analyses using time-fixed (baseline values) and time-dependent (all available measurements) methods were used to evaluate the independent relationship of RNA level and CD4 lymphocyte percent with mortality. Time-fixed models were adjusted for baseline value of the primary independent variable, age at baseline, and IVIG or placebo treatment group; time-dependent models were adjusted for all available measurements of the primary independent variable, age at time of measurement, zidovudine use at time of measurement, and IVIG or placebo treatment group.

I.3 Results Study population. Of the 376 children in the trial, 254 (68%) had 1 sample available for testing. Characteristics of the analysis cohort are shown in table 1. Children in the analysis cohort were similar to the overall study cohort in terms of sex, race or ethnicity, age at entry, entry CD4 lymphocyte count and percent, prior history of AIDS-defining infections, use of zidovudine therapy and tri-methoprim-sulfamethoxazole P. carinii pneumonia prophylaxis during the study, duration of follow-up, and percentage of mortality. Forty-one children (16%) died during the course of the trial, and an additional 51 children (20%) died during the extended follow-up period. Table 1. Characteristics of children included in the HIV-1 RNA analysis cohort (children with available baseline serum specimens) compared with characteristics of the overall National Institute of Child Health and Human Development Intravenous Immunoglobulin Clinical Trial study population.

Characteristic

Analy Overall sis cohort study cohort

vii

No. of patients % male % minority race or ethnicity % with perinatal infection Age at entry (%) 1 year 1-2 years 2-6 years >6 years Mean years (SD) Mean CD4 lymphocyte values at entry Absolute count (/mm3) % History of AIDS-defining infection at entry (%) Opportunistic infection Recurrent serious bacterial infection Opportunistic or recurrent serious bacterial infection % who used zidovudine during study % who used trimethoprimsulfamethoxazole for PCP prophylaxis (cumulative) (SD) Mean years of vital status follow-up (2.8) No. of deaths (%) Total no. of serum samples Mean no. of serum samples of patients NOTE. PCP = Pneumocystis carinii pneumonia. (36.2) (2.32)

254 55.5 92.1 90.2

376 54.8 91.8 91.5 13.6

12.2 20.1 53.1 14.6 3.41 (2.33)

19.7 52.9 13.8 3.36 1127 25.3

1105

6.4 19.1

25.1 3.9 15.7 18.1 44.1 51.2 5.1 92 1124 4.4

22.3 43.6 46.5 5.0 (2.9) 149 (39.6)

viii

Response to the vital status survey was received from 100% of trial sites. Of the 213 children in the HIV-1 RNA analysis cohort who were alive at the end of the clinical trial, 199 (93.4%) had updated vital status information available. HIV-1 RNA. A wide range of HIV-1 RNA levels were observed at baseline, from undetectable (<4000 copies/mL) to 32,000,000 copies/mL (figure 1). The median level of HIV-1 RNA at baseline was 105,000 copies/mL; the arithmetic and geometric means were 810,985 and 104,626 copies/mL, respectively. Because of the skewed distribution of RNA results, the geometric mean was considered more appropriate than the arithmetic mean, and the logarithmic (base 10) transformation was used in subsequent analyses. Overall, only 8% of children had undetectable (<4000 copies/mL) HIV-1 RNA levels at baseline and only 16% had levels < 10,000 copies/mL. Fifty percent of children had RNA levels > 100,000 copies/mL, and 22% had baseline levels > 500,000 copies/mL. Despite prolonged serum storage, the NASBA assay demon-stated a >4-fold difference in logio baseline HIV-1 RNA levels in this population. In addition, the distributions of mean and median RNA levels were similar, regardless of year of study entry, further suggesting that length of storage did not greatly affect serum HIV-1 RNA levels when measured by the NASBA assay. Most of the children entered in 1988 and 1989 (136 and 87, respectively); entry characteristics (mean age, CD4 cell percent, and history of AIDSdefining infections) were similar for children entering in each year. The 31 children who entered in 1990 had ages and CD4 cell percents similar to those of the children who entered in the previous years but were somewhat less likely to have a history of AIDS-defining infections at entry. Geometric mean baseline RNA levels were 117,175 copies/mL (logio value, 105'07) for children who entered in 1988, 88,281 (10495) for those who entered in 1989, and 102,527 (105 01) for those who entered in 1990. In a longitudinal evaluation of geometric mean HIV-1 RNA levels by age at time of measurement, RNA levels were highest among infants <1 year of age; the geometric mean was >400,000 copies/mL for infants < 12 months of age (figure 2). Although there was a slow decrease in HIV-1 RNA levels with increasing age, mean levels did not decline to < 100,000 copies/mL until after children were 3 years of age. There was a subsequent slower decline to levels below 70,000 copies/mL after children were 4-5 years of age. This pattern persisted when values from children who died or who had received zidovudine were excluded, although children who died had HIV-1 RNA at consistently higher levels.
ix

The Iog10 HIV-1 RNA copy number within subjects declined over time after study entry, with a mean slope of 0.293 (median, 0.115) Iog10 HIV-1 RNA copies/year of follow-up (95% confidence interval [CI], -0.471 to -0.115; P = .002). This pattern was unaltered by the exclusion of values from children who died or who had received zidovudine. The absolute change (final baseline values) also showed a decline (mean, -0.128; 95% CI, -0.242 to -0.013; P = .030).

Baseline HIV-1 RNA copy numbers were similar in subjects in the IVIG and placebo study arms (geometric means, 100,560 and 108,467 copies/mL, respectively; P = .78). The mean absolute change and slope of logic RNA copy number in IVIG recipients was significantly different from zero (mean absolute change, -0.24; P < .01; mean slope, -0.33; P< .01), whereas the mean change in log,0 RNA copy number in placebo recipients was not (mean change, 0.03; P = .71; mean slope, 0.26; P = .06). However, when changes in RNA levels were compared between treatment groups, there was not a significant difference between FVIG and placebo recipients in the mean change (P = .07) in or mean slope (P = .68) of HIV-1 RNA levels. Therefore, HIV-1 RNA data from both study groups were combined for most mortality analyses; in the multivari-able proportional hazards models, however, an indicator variable identifying treatment group (IVIG vs. placebo) was included as a covariate. There were 92 deaths in the analysis cohort during the trial and follow-up period. When baseline HTV-1 RNA levels were divided into categories of ^10,000, 10,001-100,000, 100,001-1,000,000, and > 1,000,000 copies/mL, a gradient of increasing mortality risk with increasing RNA level was observed (P < .001, x2 test for trend) (table 2). Mortality in children with undetectable HIV-1 RNA (<4000 copies/mL) at baseline was similar to that in children with baseline levels of 4001-50,000 copies/mL (24% and 28%, respectively), and there was a slight nonsignificant decrease in mortality rates in children with baseline levels of 50,001-100,000 copies/mL (15%). Similarly, mortality in children with HIV-1 RNA levels of 100,001-500,000 copies/mL was the same as for children with baseline RNA levels of 500,001-1,000,000 copies/mL (40% in both). Interestingly, the association of mortality with HIV-1 RNA cutpoints varied with age. For children =s2 years of age, mortality was about the same for the three lowest HIV-1 RNA categories and was elevated only when baseline HIV-1 RNA was >1,000,000 copies/mL (14%, 20%, 16%, and 74% for =s 10,000, 10,001-100,000, 100,001-1,000,000, and > 1,000,000 HTV-1 RNA copies/mL, respectively). In contrast, for children >2 years old, mortality risk increased when HTV-1 RNA exceeded 100,000 copies/mL (24%, 25%, 56%, and 67% for =slO,000, 10,001-100,000, 100,001-1,000,000^ and > 1,000,000 HIV-1 RNA copies/mL, respectively). In addition to the percentage of subjects who died during the study, the mortality rate was also examined. Table 2 provides the mortality rate (per 100 person-years) for subjects with different baseline HIV-1 RNA levels. The same gradient observed for mortality
xi

percentages was seen for mortality rates; mortality risk increased with increasing baseline HIV-1 RNA levels. The mortality rate for subjects with baseline HIV-1 RNA > 100,000 copies/mL was 2.8 times greater than the rate for subjects with HIV-1 RNA levels below this value (95% CI, 1.8-4.4; P < .001). Figure 3 A depicts a Kaplan-Meier analysis of the probability of survival by baseline HIV-1 RNA categories of^l 0,000, 10,001-100,000, 100,001-1,000,000, and > 1,000,000 copies/mL. The probability of survival during the study for children with baseline HIV-1 RNA levels < 10,000 copies/mL was similar to that observed for children with baseline levels of 10,001-100,000 copies/mL (P = .82). However, there was a significant decrease in probability of survival for those with baseline levels of 100,001-1,000,000 copies/mL compared with those with baseline levels =s 100,000 copies/mL (P = .003) and for those with baseline levels > 1,000,000 copies/ mL compared with those with baseline values =S 1,000,000 copies/mL (P < .001). The positive predictive value of different baseline HIV-1 RNA thresholds for mortality risk was evaluated (table 3). This measure reflects the percentage of patients with baseline HIV-1 RNA levels above a chosen cutpoint who subsequently died during the study or extended follow-up period. As expected, the positive predictive value increased as the HIV-1 RNA cutpoint increased. However, the positive predictive value, which is determined by the prevalence of the outcome of interest and the sensitivity and specificity of the screening test, was relatively low for the various cutpoints. Use of a baseline HIV-1 RNA level > 100,000 copies/mL as the threshold for assessing mortality risk produced a test sensitivity of 67.4% and a specificity of 59.9%. These values, combined with the overall mortality percentage of 36.2%, yielded a positive predictive value of 48.8%. The relative risk of death (RR) associated with having baseline HTV-1 RNA values above versus below selected thresholds also is shown in table 3. At 10,000 copies/mL, the RR was elevated but not statistically significant. Changing the threshold value-from 10,000 to 100,000 copies/mL increased the RR to 2.1 (95% CI, 1.4-3.0). There was only a slight increase in the RR as the baseline HIV-1 RNA threshold was raised to 1,000,000 copies/mL. HTV-1 RNA in HIV-1-Infected Children Table 2. Association of baseline HIV-1 RNA and CD4 cell percent with mortality during study and follow-up.

xii

Baselin e HIV-1 RNA level (copies/mL) 10.00 0 10.000 1-100.000 100.00 1-1.000.000 > 1.000.000 Total CD4 cell percent 25% 15%24.9% < 15%

N o. of deaths

N o. of patients

Mortalit y percent*

Person -time

Mortalit y rate

9 21 37 25 92 31 16 44 91

40 87 92 35 25 4

22.5 24.1 40.2 71.4 36.2 23.8

243.30 506.87 453.13 96.90 1300.2 1

3.70 4.14 8.17 25.80 7.08 4.07

13 0 62 60 25 2

25.8 73.3 36.1

761.10 341.12 192.18 1294.4 0

4.69 22.90 7.03

Total * P < .001, x2 test for trend, for both HIV-1 RNA level and CD4 cell percent. f Person-years of follow-up.

J Per 100 person-years of follow-up. Excludes 2 subjects who did not have an available baseline CD4 cell measurement. Subjects were grouped into quartiles on the basis of the value of their average rate of change (i.e., the slope). The mortality percentage among subjects in the fourth quartile (largest HTV-1 RNA increase) was greater than the mortality percentage in the first three
xiii

quartiles (59.6% vs. 26.9%, P < .001). Subjects also were grouped on the basis of whether their average rate of change in HIV-1 RNA level increased over the study (i.e., slope >0.0) or decreased or remained constant (i.e., slope =sO.O). The mortality percentage among those who had increasing HIV-1 RNA levels was 1.8 times greater than that among the group with decreasing or constant HIV-1 RNA levels (95% CI, 1.3-2.6; P = .001). Similar results were seen when the difference (i.e., final value baseline value) was used as a measure of intraindividual change. To further examine the relationship between HIV-1 RNA values and mortality, while controlling for potential confounding covariates, proportional hazards models were constructed. Table 4 presents the mortality risk ratios per Iog10 difference in HIV-1 RNA values. Time-fixed and time-dependent models were used that included only HIV-1 RNA (unadjusted) as well as models that controlled for study treatment group, age, and zidovudine use. The risk ratios per 1 Iog10 difference in HIV-1 RNA from these models varied from 2.2 to 3.3 and were statistically significant in each model. CD4 lymphocyte percent. - The mean baseline CD4 cell count among the analysis cohort was 1105/mm3 (SD, 892/ mm3). The mean and median baseline CD4 lymphocyte percents were 25.1% and 25.0%, respectively. The percentages of deaths and the mortality rate during the study according to the baseline CD4 lymphocyte percent is shown in table 2. Both measures indicate a strong relationship between decreasing CD4 cell percent and increasing mortality risk. Survival curves for subjects with baseline CD4 lymphocyte percents of <15% and 3=15% at baseline are shown in figure 3B. Subjects with baseline percents of <15% had a significantly poorer survival curve (P < .001). Proportional hazards models of CD4 lymphocyte percent and mortality are shown in table 4. The risk ratios presented are associated with a 5 percentage point difference in CD4 lymphocyte percents (e.g., a comparison between subjects with CD4 cell percents of 20% and 15%). There were statistically significant associations (P < .001) between lower CD4 lymphocyte percent and higher mortality risk in the time-fixed (adjusted and unadjusted) and time-dependent (adjusted and unadjusted) proportional hazards models. HIV-1 RNA and CD4 lymphocyte percent. Baseline HIV-1 RNA levels had only a borderline statistically significant correlation with baseline CD4 lymphocyte percents, as
xiv

shown in figure 4 (Pearson correlation coefficient, -0.12; P = .061). Thus, baseline CD4 lymphocyte percent does not provide an accurate indication of the specific baseline HIV-1 RNA level and vice versa. The joint relationship of HTV-1 RNA and CD4 lymphocyte percent with mortality is given in table 5. Subjects were divided into groups on the basis of the combination of their baseline HTV-1 RNA results (> 100,000 or ss 100,000 copies/mL) and CD4 lymphocyte percents (5=15% or <15%). Regardless of baseline CD4 lymphocyte percent, subjects with high baseline HTV-1 RNA levels had higher mortality than those with low HTV-1 RNA levels. Likewise, subjects with low baseline CD4 lymphocyte percents were at higher risk than those with higher CD4 lymphocyte percents, independent of baseline HTV-1 RNA level. The survival curves for the groups of subjects defined by joint baseline HIV-1 RNA level and CD4 lymphocyte percent are shown in figure 3C. The survival rates were statistically significantly different among the groups (P < .001). HTV-1 RNA Positive threshold (copies/ml, at value (%) baseline) 10,000 100,000 38.8 48.8

predictive

Relative risk* (95% confidence interval) 1.72 (0.95-3.14) 2.07 (1.44-2.96)

* Positive predictive value divided by % of deaths among patients with baseline HTV-1 RNA values below threshold. The joint relationship of HIV-1 RNA level and CD4 lymphocyte percent with mortality also was examined by proportional hazards models (table 4). Both HIV-1 RNA level and CD4 lymphocyte percent were significantly associated with mortality. The results from the joint models were similar to those seen when HTV-1 RNA levels and CD4 lymphocyte percents were modeled separately. In the time-dependent model adjusted for treatment group (IVIG or placebo), age, and zidovudine use during the study, the mortality risk ratio per Iog10 increase in HTV-1 RNA measurement was 2.75 (95% CI, 2.10-3.62; P < . 001). In the same model, the mortality risk ratio per 5 percentage point decrease in CD4 lymphocyte percent was 1.33 (95% CI, 1.20-1.47; P < .001).
xv

I.4 Discussion A wide range of HIV-1 RNA levels were observed for children in this study, from <4000 to 32,000,000 copies/mL, indicating that serum was a suitable sample type for measurement of virus load using the NASBA assay. The vast majority of these children had perinatal HIV infection, implying that the mean duration of HIV-1 infection in the cohort should be about equivalent to the mean age of the children. The baseline HIV-1 RNA level in these children, with a geometric mean value of almqst 105,000 copies/mL, was significantly higher than that reported in most infected adults with a similar 3.4-year duration of infection [3, 20]. On cross-se:ctional and longitudinal analyses, HIV-1 RNA levels were highest among the youngest children and declined with age, with the greatest decline occurring before ~2-4 years of age. These data are consistent with recent longitudinal data reported from 2 cohorts of perinatally infected children followed from birth. In these cohorts, the median peak HIV-1 RNA levels during the first few months of life ranged from 318,000 to 1,400,000 copies/mL, with a subsequent decline over the first 2 years of life to a median level of ~40,000 copies/mL [11, 12]. Thus, in contrast to the situation in infection among adults, high RNA levels persist in perinatally infected children and a set point may not be established for at least 1-3 years, possibly reflecting the lower efficiency of the developing irrimune system in containing viral replication. The poor control of viral replication observed in perinatally infected children early1 in life may be correlated with the more rapid HIV-1 disease progression observed in pediatric compared with adult HIV infection.

The average yearly decline in HIV-1 RNA levels observed in this study was 0.29 Iog10/year (median, 0.11). This value is similar to the average yearly decline (0.21 logio) reported by Mclntosh et al. [10] in a longitudinal evaluation of HIV-1 RNA in stored serum samples from a cohort of 48 infected children (median age, 29 months) from Massachusetts. These investigators found that this decline was independent of clinical, immunolpgic, and antiretroviral treatment status. Therefore, age-related changes in HIV-1 RNA levels may need to be considered'when using virus load measures to evaluate antiretroviral drug efficacy in young children.

xvi

Although there was an inverse relationship between HIV-1 RNA level arid CD4 lymphocyte percent, the correlation was only modest, 'and a 3-4 Iog10 range in number of HIV-1 RNA copies was observed for any selected CD4 lymphocyte percent. Table 4. Time-fixed and time-dependent proportional hazards models: relationship of HIV-1 RNA level, CD4 cell percent, and combined effect of HIV-1 RNA level and CD4 lymphocyte percent with mortality.

Primary independent variable(s) HlV-l RNA level CD4 cell

Time-fixed Unadjuste d 2.16 (1.67-2.81) 1.42 ed Adjust usted

Time-dependent Unadj Adjusted

2.74 2.74 (2.06-3.65) (2.16-3.47) 1.42 (1.29- 1.54(1.39-

3.32 (2.54-4.33) 1.52

NOTE. Data are mortality risk ratio per 1 logic increase in HIV-1 RNA level and per 5 percentage point decrease in CD4 lymphocyte percent (95% confidence interval). Timefixed, unadjusted model includes only baseline value of primary independent variable(s); time-fixed, adjusted model includes baseline value of primary independent variable), plus treatment group (intravenous immunoglobulin or placebo) and age at baseline; timedependent, unadjusted model includes all available measurements of primary independent variable(s); time-dependent, adjusted model includes all available measurements of primary independent variable(s), plus treatment group, age at time of measurement, and zidovudine use at time of measurement. P < .001 for all comparisons. This poor correlation of CD4 lymphocyte count with viral RNA load has also been observed in cohorts of infected adults [5, 21]. The prolonged follow-up (3=5 years) available in this study enables evaluation of the association of HIV RNA levels and CD4 cell percents with long-term clinical outcome that is unavailable in any published pediatric cohort data. Higher baseline HIV-1 RNA levels were associated with increased long-term mortality risk. This is similar to what has been observed among infected adults [1, 3, 5]. However, defining a discrete HIV-1 RNA cutpoint for potential clinical decision-making is difficult, as evidenced by the poor predictive value of
xvii

HIV-1 RNA levels < 1,000,000 copies/mL for mortality risk. The HIV-1 RNA threshold of 10,000 copies/mL suggested as a cutoff for considering initiation of therapy for infected adults [7] had a very poor predictive value for mortality among this cohort of infected children, with a positive predictive value of only 39%. When a threshold of 100,000 copies/mL was used, a 2-fold RR was seen between subjects above this cutpoint compared with that for subjects below this cutpoint. The positive predictive value, however, was still only 49%. Additionally, because of age-related changes in HIV-1 RNA early hi life, the prognostic value of specific RNA levels may differ by age.

In studies of infected adults, an increase in HIV-1 RNA levels over time has been associated with elevated mortality risk, and therapy-related declines in RNA levels have been associated with improved prognosis [5, 6, 22, 23]. In this pediatric cohort, there also was an association between increases in HIV-1 RNA levels observed during the clinical trial and elevated long-term mortality risk. It is important to note that this pediatric clinical trial was conducted during a time when few antiretroviral therapies were available for children.

Table 5.

xviii

Association of baseline HIV-1 RNA level and CD4 lymphocyte percent with mortality during study.

HIV-1 RNA CD4 level(copies/mL) cell percent 100,000 > 100,000 100,000 > 100,000 15 15 <15 <15

No. No. of deaths patients 15 32 15 29 103 89 24 36

of

Morta lity (%) 14.6 36.0 62.5 80.6

Zidovudine became available for pediatric use only after the study had completed most of its enrollment. Although the time-dependent multivariate model incorporated zidovudine therapy as a covariate, the trial was not designed to evaluate antiretroviral therapy, and thus the effect of antiretroviral therapy and prognostic value of changes hi RNA levels with therapy other than zidovudine could not be assessed. Finally, as observed among infected adults [5, 6, 24], baseline HIV-1 RNA level and baseline CD4 lymphocyte percent were independently predictive of mortality risk. Regardless of whether baseline HIV-1 RNA levels were above or below 100,000 copies/mL, the mortality percentage for subjects with baseline CD4 lymphocyte percents below 15% was greater than the mortality percentage among subjects with CD4 lymphocyte values above 15%. In fact, by some measures, the prognostic value of the baseline CD4 lymphocyte percent was greater than the prognostic value of HIV-1 RNA levels. For example, the sensitivity, specificity, and positive predictive value of a baseline HIV-1 RNA level > 100,000 copies/mL was 67.4%, 59.9%, and 48.8%, respectively. By comparison, the sensitivity, specificity, and positive predictive value of a baseline CD4 lymphocyte percent of <15% was 48.3%, 90.1%, and 73.3%, respectively. Also, HIV-1 RNA level and CD4 lymphocyte percent variables were each statistically significant in proportional hazards models that included both markers. The relationship between HIV-1 RNA level and CD4 lymphocyte percent, as well as their prognostic values, may be affected by the pathogenesis of HIV infection. An HIV-1 RNA measurement obtained shortly after infection (but after the initial burst of viremia) may be more predictive of disease progression than a CD4
xix

lymphocyte measurement obtained at the same time. After a subject's immune system has had time to be affected by the infection, however, immunologic measures are likely to become more relevant. This study group was first examined an average of almost 3.5 years after infection, and both HIV-1 RNA and CD4 lymphocyte levels were independent and complementary markers of disease stage. In many instances, a clinician may not examine a patient until several years after the initial infection. Therefore, both markers should be considered together for decision-making regarding therapy and evaluation of response to antiretroviral agents.

xx

References
1. Katzenstein TL, Pedersen C, Nielsen C, Lundregren JD, Jakobsen PH, Gersetof J. Longitudinal serum HIV RNA quantification : correlation to viral phenotype at seroconversion and clinical outcome. AIDS 1996; 10 :167-73 2. Henrard DR, Philips JF, Muenz LR, et al. Natural history of HIV-1 cell-free viremia JAMA 1995;274:554-8 3. Havlir DV, Rihman FF. Viral dynamics of HIV : implications for drug developmeny and therapeutic strategies. Ann Intern Med 1996;124:984-94

xxi