Natural History of Human Immunodeficiency Virus Type 1 Viremia after Seroconversion and Proximal to AIDS in a Large Cohort of Homosexual Men

Abstract

The natural history of human immunodeficiency virus type 1 (HIV-1) viremia and its association with clinical outcomes after seroconversion was characterized in a cohort of homosexual men. HIV-1 RNA was measured by reverse-transcription polymerase chain reaction (RT-PCR) in stored longitudinal plasma samples from 269 seroconverters. Subjects were generally antiretroviral drug naive for the first 3 years after seroconversion. The decline in CD4 lymphocyte counts was strongly associated with initial HIV RNA measurements. Both initial HIV RNA levels and slopes were associated with AIDS-free times. Median slopes were +0.18, +0.09, and −0.01 log₁₀ copies/mL, respectively, for subjects developing AIDS <3, 3–7, and >7 years after seroconversion. In contrast, HIV RNA slopes in the 3 years preceding AIDS and HIV RNA levels at AIDS diagnosis showed little variation according to total AIDS-free time. HIV RNA load at the first HIV-seropositive visit (∼3 months after seroconversion) was highly predictive of AIDS, and subsequent HIV RNA measurements showed even better prognostic discrimination.

The remarkable prognostic value of a single measurement of plasma virus load among patients infected with human immunodeficiency virus type 1 (HIV-1) is widely recognized. Patient groups for whom this value has been demonstrated include homosexual men [1], women [2], injection drug users [3], and hemophiliacs [4]. Recent studies involving serial measurements of HIV RNA load have fostered various opinions for and against a viral “set point” theory, which posits that virus load for an infected patient tends to stabilize relatively soon after HIV seroconversion, with the “set point” being a strong indicator of future disease progression [5–10]. Researchers have also hypothesized that an early dynamic between viral replication and the immune response may influence a person's longitudinal pattern of immunologic and virologic markers, as well as the prognosis for clinical progression [5].

Despite a growing number of studies involving longitudinal monitoring of HIV RNA load, few have focused primarily on HIV-1 seroconverters. Of those that have, the study populations have often been limited in size [5, 6]. In a recent study [11], however, investigators measured HIV RNA over time for 149 Italian seroconverters from multiple risk groups and documented an association between long-term HIV RNA slopes and clinical progression.

In the present study, longitudinal HIV RNA measurements from 269 homosexual male seroconverters were obtained from stored plasma specimens from a national repository maintained for the Multicenter AIDS Cohort Study (MACS). Objectives included documentation of HIV RNA patterns for periods up to 10 years after seroconversion, with particular interest in untreated natural history within the first few years of seroconversion, the relationship between CD4 cell count before HIV infection and HIV RNA load after infection, and patterns of HIV RNA near the time of the initial AIDS diagnosis. This focus, in addition to the large size of the cohort, distinguishes the current study from previous reports. We provide data on variation in HIV RNA trajectories early and late in disease progression, according to overall AIDS-free time since seroconversion. By taking advantage of the longitudinal measurements and documented dates of last seronegative and first HIV-seropositive visits, we investigate possible variation in the prognostic value of plasma HIV RNA load relatively early in infection, according to the time since seroconversion. To explore the possibility of an early HIV RNA-CD4 cell dynamic in this cohort, we also relate CD4 cell counts obtained before seroconversion with plasma HIV RNA loads and CD4 cell counts at the first HIV-seropositive visit.

Subjects and Methods

Study population and laboratory measurements

The MACS is an ongoing cohort study in which 3427 of the 5622 AIDS-free homosexual/bisexual men enrolled were seronegative at baseline, and 511 were observed to seroconvert before the date of analysis for this study (30 September 1996). Participants were primarily white (88%), and their ages at enrollment ranged from 18 to 60 years. Semiannual visits are ongoing at centers in 4 US cities (Baltimore, Chicago, Los Angeles, and Pittsburgh) and routinely consist of physical examinations, the collection of laboratory samples, and the administration of detailed questionnaires. Men who were not HIV-positive at study entry were tested for seroconversion semiannually by ELISA, with confirmation of positive tests by Western blotting. Details about the composition of the cohort, surveillance for HIV- and AIDS-related outcomes, and the demographic and laboratory data collected appear elsewhere [12, 13].

Among the persons who were seronegative at enrollment into the MACS and who were observed to seroconvert during follow-up, we identified for this study a group comprising those who had CD4 cell counts available from both their last HIV-seronegative (LN) and first HIV-seropositive (FP) visits and for whom the dates of these visits were <8 months apart. On average, the LN and FP visits took place at −3 and +3 months from seroconversion. We required subjects to have at least 2 plasma specimens after seroconversion available in the national repository. The number of men fulfilling these criteria was 269, and the total number of plasma samples in which HIV RNA was quantified was 2527, covering a time span of 10 years after seroconversion. For the majority of these men (90%), the LN and FP visit dates were <7 months apart. By study design, all available plasma samples at the LN and FP visits, and for 2 years thereafter, were assayed for HIV RNA. Beyond 2 years after the FP visit, 1 sample per year from each participant was selected for assay, for a total of up to 10 years of follow-up on each subject.

At each semiannual MACS visit, absolute CD4 T lymphocyte counts were obtained via established flow cytometric procedures [14]. HIV RNA was measured in stored samples by reverse-transcription polymerase chain reaction (Amplicor; Roche Diagnostics, Nutley, NJ), with an assay detection limit of 400 copies/mL. Previous investigators have provided details regarding the reliability of this assay and the appropriateness of stored repository specimens for such studies [15, 16].

Statistical analysis

We applied multiple linear regression to characterize the association between the seroconversion mean CD4 cell count and the log-transformed HIV RNA load at the FP visit, after first normalizing CD4 cell counts by a 4th-root transformation [17]. To further describe overall trends in HIV RNA load and CD4 cell count over time, we fitted random-effects linear models by use of the SAS mixed procedure (SAS Institute, Cary, NC) [18]. These models provide estimates of the average linear marker trajectories over time, while accounting for correlations among repeated measurements from the same subjects. Confidence bands for the resulting regression lines were calculated by use of an adaptation of the Scheffé procedure [19]. Since only a small percentage (∼3%) of samples had undetectable (<400 copies/mL) virus loads, it was deemed reasonable to impute a value of 300 copies/mL for these samples.

To assess differences in longitudinal patterns of HIV RNA load according to time to clinical AIDS (i.e., not including the definition based only on a low CD4 cell count) [20], we considered 4 subgroups according to time to AIDS: AIDS occurring <3 years after the FP visit, in 3–7 years, or in >7 years or AIDS-free as of the date of analysis with at least 9 years of follow-up. We fitted simple linear regressions of HIV RNA loads expressed in log₁₀ copies/mL and of CD4 cell counts over time to the data from each participant and summarized the results separately for each of the 4 subgroups.

We constructed Kaplan-Meier curves to assess the prognostic value of HIV RNA load for the development of AIDS, on the basis of categories defined by cut points at 10,000, 25,000, and 75,000 copies/mL. Log rank tests and proportional hazards regression models [21] were applied to estimate and discriminate risk across the 4 HIV RNA load categories (<10,000, 10,000–24,999, 25,000–75,000, and >75,000). We performed the latter analysis separately by use of the HIV RNA quantified at (and measuring follow-up time from) different time points relative to seroconversion. Namely, these time points were the FP visit that occurred ∼3 months after seroconversion and each of the 3 subsequent semiannual visits (∼9, 15, and 21 months after seroconversion).

Results

The median age at the FP visit for the 269 participants was 33.9 years (interquartile range, 28.7–39.5 years), with 84% of the men being white and >50% having college degrees. The median time elapsed between the LN and FP visits was 6.0 months (interquartile range, 5.7–6.4 months).

Preseroconversion HIV RNA loads and CD4 cell counts

Of the 269 subjects studied, 40 (15%) had detectable HIV RNA at the LN visit (∼3 months before seroconversion). Among those 40 men, the mean change in HIV RNA load between the LN and FP visits was −0.35 log₁₀ copies/mL. Although not statistically significant (P = .11), this may suggest that between the LN and FP visits, these men were in the declining phase of viremia that occurs soon after HIV infection.

A total of 1148 CD4 cell count measurements were obtained from 259 men before seroconversion. The average of the preseroconversion mean CD4 cell counts for these men was 988 cells/mm³, and the average decline from the preseroconversion mean CD4 cell count to the mean CD4 cell count at the FP visit was 221 cells/mm³. There were significant correlations between the magnitude of this decline and both the preseroconversion mean CD4 cell count (Spearman's r = .56, P < .001) and the HIV RNA level at the FP visit (Spearman's r = .28, P < .001). Univariately, the preseroconversion mean CD4 cell counts showed a weak trend toward a positive association with the level of HIV RNA at the FP visit (Spearman's r = .12, P = .064). In a linear model in which HIV RNA in log₁₀ copies/mL was regressed simultaneously on the 4th-root-transformed CD4 cell count at the FP visit and the individual-specific mean of the 4th-root-transformed preseroconversion CD4 cell counts, there was a strong positive association between the level of HIV RNA and this preseroconversion mean CD4 cell count (P < .001). This strong association was present with or without applying the normalizing transformation to CD4 cell counts. Thus, when the CD4 cell count at the FP visit was controlled for, those with higher average CD4 cell counts before seroconversion had significantly higher plasma HIV RNA levels at the FP visit.

HIV RNA load and CD4 cell count trajectories in the first 3 years after seroconversion

Table 1 provides descriptive statistics (medians, interquartile ranges, and correlations) for the HIV RNA load and CD4 cell count at the FP visit (∼3 months after seroconversion) and the 6 ensuing semiannual visits. To focus on untreated natural history, only data from visits before 1 January 1990 were included. As a result of this restriction, 218 (instead of the full 269) subjects contributed the data presented in table 1, and the numbers of men contributing at each time point declined, because 57 of the 218 subjects serocon-verted <3 years before 1 January 1990. There were also 21 deaths and 4 losses to follow-up among seroconverters before 1 January 1990. Six months after the FP visit (∼9 months after seroconversion), only 1.8% of these subjects reported any use of antiretroviral therapy. This percentage remained small throughout the period summarized in table 1, increasing to 9.1% at the last visit considered (3 years after the FP visit). There was essentially no use of combination therapy (i.e., ⩾2 nucleoside reverse-transcriptase inhibitors) during this period (<1% at the last visit). The data in table 1 indicate relative stability in HIV RNA loads for the group as a whole over the first 2 years after the FP visit, with some indication of an increase thereafter. In contrast, the decline in CD4 cell counts over the same period is much more striking.

Table 1

Summary of human immunodeficiency virus (HIV) RNA load and CD4 cell counts during the first 3 years after the first HIV-seropositive (FP) visit.

Figure 1 depicts a scatter plot of the data summarized in table 1, together with the estimated average linear trajectories of HIV RNA loads in log₁₀ copies/mL and CD4 cell counts, with ∼95% confidence bands. The average rate of change (slope) in HIV RNA load (left) was +0.029 (SE = 0.024) log₁₀ copies/mL per year, corresponding to a non-statistically significant 7% increase in virus load per year (P = .22). In contrast, the model fitted to the CD4 cell count data (right) suggests a highly significant decline of 109 cells/mm³/year (SE = 9.53; P < .001) over the same time period.

Figure 1

Average trajectories of human immunodeficiency virus (HIV) RNA load (log₁₀ scale) and CD4 cell counts during the first 3 years after the first HIV-seropositive (FP) visit (at ∼3 months after seroconversion) for 218 participants in the Multicenter AIDS Cohort Study. Only measurements before 1 January 1990 are included, to avoid potential confounding effects of antiretroviral therapy.

Levels and slopes of HIV RNA loads and CD4 cell counts: progression to AIDS

To focus on early changes in HIV RNA loads and CD4 cell counts according to disease progression, linear regression models were fitted to the data from each subject from 4 subgroups defined according to time to AIDS. Table 2 summarizes the initial levels and slopes of both markers for each of the 4 subgroups. A total of 173 men contributed data to the table. Twelve men who seroconverted after 1 January 1990 were excluded, as were 19 men who developed AIDS but for whom the date of diagnosis was not ascertainable within 6 months and 65 men who never developed AIDS but were followed up for <9 years. Across the subgroups, there were clear gradients in both the early levels and slopes of HIV RNA load and of CD4 cell count.

Table 2

Changes in human immunodeficiency viras (HIV) RNA load (log₁₀ copies/mL) and CD4 cell counts (cells/mm³) over the first 3 years after the first HIV-seropositive (FP) visit, by time-to-AIDS.

Inferences about whether the early levels and slopes of HIV RNA differed across the 4 subgroups in table 2 were undertaken via random-effects linear models. Figure 2 illustrates statistically significant differences across the time-to-AIDS subgroups, both in the initial level of log-transformed HIV RNA (P < .001) and in the slope over the first 3 years after the FP visit (P = .013). Subgroup differences in initial CD4 cell counts (P = .020) and slopes (P < .001) were also significant (data not shown), with most of the slope difference being attributable to a precipitous decline in CD4 cell counts for men developing AIDS in <3 years.

Figure 2

Average trajectories of human immunodeficiency virus (HIV) RNA load (log₁₀ scale) during the first 3 years after the first HIV-seropositive (FP) visit (∼3 months after seroconversion) in the Multicenter AIDS Cohort Study for 4 subgroups defined by different time intervals to development of AIDS (A-D). Only measurements before 1 January 1990 are included, to avoid effects of antiretroviral therapy.

Figure 3 provides box plots of HIV RNA load by year for each of the 4 time-to-AIDS subgroups. The number of men contributing data during each year is listed at the top of each panel. A subject could contribute up to 2 measurements during each year. In figure 3a–3c, time is measured backward starting at the initial AIDS diagnosis, and in figure 3d, time is measured forward from seroconversion. Figure 3 shows distinct patterns of HIV RNA load according to prognosis, with high initial levels followed by an increase, characterizing rapid progression to AIDS (figure 3a), and low initial levels that remained remarkably stable, characterizing nonprogression (figure 3d). The subgroup that developed AIDS in 3–7 years showed moderate initial HIV RNA levels followed by a consistent increase (figure 3b), whereas the subgroup developing AIDS in >7 years showed similar initial levels but an early period of relative stability before HIV RNA levels increased (figure 3c).

Figure 3

Box plots depicting the distribution of human immunodeficiency virus (HIV) RNA load (log₁₀ scale) in 4 subgroups (A-D) of HIV-1 seroconverters, defined by time to AIDS development. Nos. of subjects contributing data during each year are given at top.

HIV RNA trajectories near the time of AIDS

Table 3 summarizes HIV RNA levels and CD4 cell counts by year for the 3 years preceding the date of the initial diagnosis of AIDS. Linear random-effects models were applied to all HIV RNA data available within 3 years of the date of diagnosis for the 3 groups that developed AIDS in <3, 3–7, and >7 years after the FP visit, respectively. Estimates and confidence bands were obtained for the slope during this time and for the level of HIV RNA at the time of AIDS. Figure 4(left) shows estimates and confidence bands for the HIV RNA levels at the FP visit (corresponding to the intercepts of the regression lines in figure 2) and for the levels at AIDS, by time-to-AIDS subgroups. Although the levels at the FP visit were significantly different, as noted previously (P < .001), the levels at AIDS were not (P = .65). Figure 4(right) gives a similar representation for the HIV RNA slopes within 3 years of the FP visit (corresponding to the trajectories in figure 2) and for the slopes within 3 years of progression to AIDS. Again, the statistically significant differences across subgroups noted within the first 3 years of seroconversion (P = .013) are no longer evident late in progression (P = .71).

Table 3

Summary of human immunodeficiency viras (HIV) RNA loads (copies/mL) and CD4 cell counts (cells/mm³) during the 3 years preceding the initial diagnosis of AIDS.

Figure 4

Estimates and 95% confidence intervals for levels and slopes of human immunodeficiency virus (HIV) RNA load (log₁₀ scale). Initial levels at the first HIV-seropositive (FP) visit at ∼3 months after seroconversion and slopes within the subsequent 3 years (◯) and levels at the time of the initial AIDS diagnosis and slopes for the 3 years preceding AIDS (A) are shown.

Early prognostic value of HIV RNA load by time since seroconversion

Figure 5a shows Kaplan-Meier curves for AIDS-free survival for the 249 participants whose date of AIDS diagnosis was known within 6 months, according to the 4 categories of HIV RNA load at the FP visit defined by cut points of 10,000, 25,000, and 75,000 copies/mL. Figure 5b–5d provides similar representations, but with HIV RNA measured at the 3 subsequent semiannual visits that occurred ∼9, 15, and 21 months after seroconversion and survival time measured from those respective visits. Only 5 subjects developed AIDS within 1.5 years of their FP visit, so there is little concern about survival bias in interpreting figure 5. Relative hazard estimates are given, with the category of <10,000 copies/mL as reference. Log rank tests were highly significant in each case (P < .001 for each). Although HIV RNA load is a strong predictor at all 4 time points, there is a gradient in the log rank statistics that corresponds well with the relative hazard estimates and the actual survival curves, suggesting that prognostic discrimination provided by HIV RNA load improves when it is measured at later times after seroconversion. In particular, the middle 2 HIV RNA load categories correspond to very similar risk at the FP visit but are much more distinct thereafter. The predictive value of HIV RNA load had stabilized by the later 2 visits, ∼1 and 1.5 years after the FP visit.

Figure 5

Kaplan-Meier curves illustrating AIDS-free survival for 4 categories of human immunodeficiency virus (HIV) RNA loads (in thousands [k] of copies/mL). Separate plots are shown with HIV RNA load measured and time anchored at the first HIV-seropositive (FP) visit at ∼3 months after seroconversion and at each of 3 subsequent semiannual visits at ∼9 (FP+1), 15 (FP+2), and 21 months (FP+3) after seroconversion (A-D, respectively). RH, relative hazard estimate.

Discussion

This study examined the natural history of HIV-1 viremia within the first 3 years after seroconversion in a population of essentially antiretroviral drug-naive persons and documented patterns of HIV load over extended follow-up, with emphasis on the years immediately preceding the diagnosis of AIDS. The findings pertain to a well-studied cohort of homosexual men and contribute to the understanding of HIV pathogenesis. They supplement information from other recent longitudinal studies among various risk groups, including a study of Italian seroconverters [11] and a study documenting differences in HIV RNA slopes among rapid progressors and controls by sex in a predominantly injection drug-using cohort [22].

The small and non-statistically significant average increase in HIV RNA (+0.03 log₁₀ copies/mL per year) over the first few years after seroconversion is consistent with the findings of other recent studies. However, we found this overall stability of virus load to be the result of a combination of distinctly different patterns according to the rate of progression to clinical AIDS. Men with different AIDS-free times were characterized by clear variation in initial RNA levels as well as differences in 3-year slopes. Those not developing AIDS displayed low early HIV RNA levels and stability in HIV RNA load over time. Other subgroups had higher initial RNA levels and showed increases that were greater according to rapidity of progression (figure 2). The consistency and magnitude of disparity in these patterns across subgroups argue against a blanket concept of a viral steady-state or set point, if we take this term to imply that RNA levels remain constant for an extended period after seroconversion.

This is the first study to contrast variation in early HIV RNA trajectories for subgroups defined by AIDS-free survival, with the tendency toward invariance in these trajectories as AIDS becomes imminent. For 3 time-to-AIDS subgroups (<3, 3–7, and >7 years) in our study, the slope during the 3 years immediately preceding progression to AIDS was 0.14–0.20 log₁₀ copies/mL per year, and the estimated level at the time of progression to AIDS was 5.2–5.3 log₁₀ copies/mL. This consistency is in marked contrast to the early patterns of viremia seen in these subgroups (figure 4), and it adds new information to recent and ongoing investigations into viral dynamics as HIV disease progresses [23]. Although there was some use of monotherapy or combination antiretroviral therapy among the participants before progression to AIDS, there was no exposure to potent therapy with protease inhibitors that would be expected to markedly alter clinical or virologic progression [24].

Given the semiannual nature of follow-up visits in the MACS, documentation of the early HIV RNA “peak” was not expected in this study. Nevertheless, HIV RNA was detected at the last seronegative visit (an average of 3 months before seroconversion) in 15% of seroconverters. The slight decrease (−0.35 log; P = .11) in HIV RNA levels from the LN to the FP visit among the 40 subjects with detectable virus before documented seroconversion provides some evidence of earlier peaks in viremia. However, the insignificant magnitude of this decline served to allay concerns about the impact of early peaks on the primary analysis [6]. For 217 men who overlapped with a prior study in the MACS exploring methods for early detection of virus [25], we obtained strong agreement regarding the ability to detect virus at the LN visit (agreement for 208 of 217 men; κ statistic = 0.86).

The initial period of HIV infection in this cohort was characterized by a familiar steep decline in CD4 cell counts after seroconversion, with generally more severe decline among those with higher CD4 cell levels before infection. We also found a positive correlation between preseroconversion mean CD4 cell counts and HIV RNA loads at the FP visit (3 months after seroconversion on average), which was highly significant after adjusting for the CD4 cell count at the FP visit (P < .001). This suggests that the more severe drop in CD4 cell count among those with higher counts before infection may well reflect greater HIV replication rather than solely representing an intrinsic but uninteresting regression to the mean effect. Such greater HIV replication is qualitatively consistent with findings of prior studies that have identified a “predator-prey” dynamic, in which the greater number of target (CD4) cells available after initiation of antiretroviral therapy may lead to a rebound in HIV RNA to levels above baseline when therapy fails or is discontinued [26]. More detailed study of this early dynamic could contribute to insights about virologic and immunologie interactions relevant to early prognosis [5, 26, 27].

Importantly, this study provides historical control data about CD4 cell counts and HIV RNA levels and slopes in the first few years after HIV infection. This information could be of importance for the evaluation of future HIV vaccines aimed not at preventing infection but at reducing the level of viremia and risk of disease progression following seroconversion [10].

Multicenter AIDS Cohort Study Investigators

Multicenter AIDS Cohort Study investigators include the following: Johns Hopkins University School of Hygiene and Public Health, Baltimore: Joseph Margolick (principal investigator), Haroutune Armenian, Homayoon Farzadegan, Nancy Kass, Justin McArthur, Steffanie Strathdee, and Ellen Taylor; Howard Brown Health Center and Northwestern University Medical School, Chicago: John Phair (principal investigator), Joan Chmiel, Bruce Cohen, Maurice O'Gorman, Daina Variakojis, Jerry Wesch, and Steven M. Wolinsky; UCLA Schools of Public Health and Medicine, Los Angeles: Roger Detels (principal investigator), Barbara Visscher, Janice Dudley, John Fahey, Janis Giorgi, Andrew Kaplan, Oto Martinez-Maza, Eric Miller, Hal Morgenstern, Parunag Nishanian, John Oishi, Jeremy Taylor, and Harry Vinters; University of Pittsburgh Graduate School of Public Health, Pittsburgh: Charles R. Rinaldo (principal investigator), James Becker, Phalguni Gupta, Lawrence Kingsley, John Mellors, Sharon Riddler, and Anthony Silvestri; Data Coordinating Center, Johns Hopkins University School of Hygiene and Public Health, Baltimore: Alvaro Muñoz (principal investigator), Linda Ahdieh, Stephen Gange, Lisa Jacobson, Cynthia Kleeberger, Robert Lyles, Steven Piantadosi, Ellen Smit, Sol Su, Patrick Tarwater, and Traci Yamashita; National Institutes of Health, Bethesda, Maryland: National Institute of Allergy and Infectious Diseases, Carl Dieffenbach (project officer); National Cancer Institute, Sandra Melnick.

Acknowledgments

We thank the dedicated participants and staff of the Multicenter AIDS Cohort Study.

Presented in part: 6th Conference on Retroviruses and Opportunistic Infections, Chicago, February 1999.

Informed consent was obtained from study participants, and human experimentation guidelines of the respective institutions conducting the study were followed.

Financial support: National Institutes of Health (cooperative agreements AI-35042, -35043, -37984, -35039, -35040, -37613, and -35041).

Author notes

a

Present affiliation: Department of Biostatistics, Rollins School of Public Health of Emory University, Atlanta, Georgia.

b

Multicenter AIDS Cohort Study investigators are listed after text.