INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

On the basis of numerous studies recently showing the predictive value of human immunodeficiency virus (HIV) type 1 (HIV-1) load on disease progression (9, 13, 14, 17, 26), viral loads are currently used in combination with CD4⁺ cell count to estimate the stage of disease and guide therapeutic decisions. Most studies of viral load have been based on viral loads in white homosexual men (HM) (13, 14), African-American injection drug users (26), or hemophiliacs (16). Studies which have evaluated viral load among heterogeneous populations are sparse. One study which included multiple risk groups but which consisted of predominantly white HM suggested that higher viral loads exist among males, among HM, and among non-drug users (9).

Use of the total number of copies of HIV-1 RNA per ml of plasma to measure viral burden includes all viral RNA particles regardless of the level of infectivity. In contrast, the cell-associated infectious HIV-1 load, measured by the quantitative microculture assay, measures biologically functional and infectious cell-associated virus, i.e., the amount of cell-associated HIV-1 capable of infecting donor cells from an uninfected person by a coculture technique. Two recent studies have compared the two assays and showed the correlation to range from 0.52 to 0.54 (10, 18). These studies mostly consisted of white HM, and it is unclear whether these two virologic measurements correlate equally among the different risk and gender groups.

For these reasons we compared the levels of HIV-1 RNA in the plasma and the cell-associated infectious HIV-1 loads in the peripheral blood between HIV-1-infected male and female injection drug users (IDUs) and HM, while at the same time we evaluated the relationship between these two virologic measures.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Study population. Participants in this study were IDUs in the Baltimore, Maryland-based AIDS Link to Intravenous Experiences (ALIVE) study or HM in the Study to Help the AIDS Research Effort (SHARE) study, which is the Baltimore site of the Multicenter AIDS Cohort Study. Both cohorts were recruited to study the natural history of HIV disease and to screen for new HIV infections. The designs of these cohort studies have been described elsewhere (8, 25). The ALIVE participants were actively recruited through community outreach programs between February 1988 and March 1989, whereas the Multicenter AIDS Cohort Study-SHARE participants were recruited in 1984. The IDUs were predominantly black individuals of lower socioeconomic status who were actively injecting drugs (25), whereas the HM were predominantly white individuals of middle to upper socioeconomic status (8). All were required to be 18 years of age, to be AIDS free at entry, and to consent to participation. In addition, IDUs were required to have a history of injection drug use since 1977.

Data collection. During the regular semiannual follow-up visits in the respective outpatient clinics, the HIV-infected participants in both study populations underwent interviews and physical examination and had blood drawn for T-cell subset studies. Additional aliquots of plasma were stored in heparinized tubes at 70°C for future studies. The data collected included detailed information on demographics, medical history, illicit drug use, and sex practices during the previous 6 months. Separate consents were obtained to secure the release of medical information. An additional 10 ml of heparinized blood was drawn for cell-associated infectious HIV-1 load quantification at each visit during the 2-year recruitment period of this substudy. Only the first infectious HIV-1 load measurement was considered in this analysis. Plasma HIV RNA levels were later quantified at the same visit as the initial infectious HIV load measurement, when frozen plasma was available.

Laboratory methods. Antibodies to HIV-1 were measured with a commercially available enzyme-linked immunosorbent assay kit (Genetic Systems, Seattle, Wash.), and the results for repeatedly positive specimens were confirmed by Western blotting (Dupont, Wilmington, Del.). Measurement of T-cell subsets was performed in one laboratory by flow cytometry according to a whole-blood staining method, which has been described previously (7, 12), and absolute counts were determined by obtaining an automated complete blood count and differential.

Statistical methods. The analysis described here was limited to those participants for whom both viral load measurements were available at the same visit. Standard summary measures were presented for both categorical and continuous variables. Univariate comparisons by population, as defined by gender and risk group, were made separately for each measure of viral load and other characteristics by nonparametric analysis of variance. Comparisons were also made within the following three CD4⁺ cell count categories: <200, 200 to 499, and 500 cells/µl. Linear regression models were used to compare viral loads by risk group while controlling for other potential confounders. To normalize the distributions both viral load measurements were transformed on the log₁₀ scale prior to regression analyses. Observations falling below the lower detectable limit of the assay were recoded to one-half of the lower limit prior to the transformation (0.2 for the QMC assay and 200 for the RT-PCR assay). Two observations which had extremely influential CD4⁺ cell counts were Winsorized to the 99th percentile (1,342 cells/µl) (21).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

A total of 547 subjects (299 IDUs and 248 HM) had a cell-associated infectious HIV load measurement during the recruitment period. Of those, 485 (89%) had sufficient frozen plasma available for the HIV RNA load quantification (performed in 1997) at the same visit as the infectious viral load quantification. The current analysis was restricted to those 485 subjects (190 male IDUs, 92 female IDUs, and 203 HM). Comparing subjects with plasma available versus those without plasma available revealed that those excluded were mostly HM who were more likely to have had an AIDS diagnosis. This was inherent in the SHARE study design, because those with AIDS were not required to provide as much blood for repository storage.

Most IDUs were African American (96%), whereas only 20% of the HM were African American (Table 1). Slightly more than one-third of the participants had CD4⁺ cell counts of 500 cells/µl, while almost one-quarter had <200 cells/µl. Few subjects had previously developed AIDS (7%), while 4% were considered recent seroconverters (estimated seroconversion date within the last 6 months). The IDUs consisted of more recent seroconverters and slightly fewer subjects with a previous AIDS diagnosis than HM. While more than half of the HM reported that they had used any antiretroviral therapy (zidovudine, dideoxyinosine, dideoxycytosine, or stavudine) during the prior 6 months, fewer than 30% of the IDUs reported that they had used any antiretroviral therapy. The greatest proportion of nondetectable infectious HIV-1 loads were observed among female IDUs (26%), followed by male IDUs (18%) and HM (11%). Fewer subjects (5%) had nondetectable HIV RNA loads, with similar proportions across groups.

The median ages differed between the three groups (P < 0.001) (Table 2). Although the CD4⁺ cell counts did not differ by group (P = 0.525), median levels of both infectious HIV (P < 0.001) and HIV RNA (P = 0.036) were statistically different between the three groups. The median infectious HIV load for HM (16.2 IUPM) was twice that for the male IDUs (8.0 IUPM) and almost three times that for the female IDUs (5.5 IUPM). For HIV RNA load, HM and male IDUs had similar loads (94,563 versus 94,557 copies/ml, respectively), while the female IDUs had a median load roughly a quarter of a log lower (51,522 copies/ml). Despite similarities in CD4⁺ cell counts, there were significant differences in median percent CD4⁺ cells by group (P < 0.001). After stratification by CD4⁺ cell count category, significant differences by group still remained for age and percent CD4⁺ cells within each category, for infectious HIV load and HIV RNA load within the categories of <200 and 500 cells/µl, and for CD4⁺ cell count within the category of <200 cells/µl.

The CD4⁺-adjusted infectious HIV load was estimated to be 0.20 log₁₀ lower among female IDUs than male IDUs (P = 0.100) and 0.25 log₁₀ higher among HM than male IDUs (P = 0.009) (Table 3). The corresponding estimated differences in HIV RNA load were 0.14 log₁₀ lower among female IDUs (P = 0.17) and 0.08 log₁₀ higher among HM (P = 0.29) compared to those in male IDUs (Table 3). Age, AIDS status, and recent antiretroviral therapy were also considered and were not found to be associated with either viral load measure (data not shown). All comparisons were essentially unaltered after controlling for recent HIV seroconverter status (data not shown) or, alternatively, excluding recent HIV seroconverters (Table 3).

Figure 1 graphically displays the comparisons of viral load made in Table 3 (excluding data for the 20 recent HIV seroconverters) by presenting the estimated regression lines for each population. Regarding the level of infectious virus (Fig. 1A), formal tests for interaction indicated a common slope for all three groups (P > 0.5 for each comparison). As observed in Fig. 1A and previously presented in model 1 (Table 3), estimated levels of infectious virus are higher among HM and lower among female IDUs relative to those among male IDUs. When combining data for the groups, females had an estimated one-third decrease in log₁₀ infectious viral load relative to those among males (P = 0.004) and HM had a higher level compared to those among IDUs ( = +0.29; P = 0.001).

View larger version (21K): [in this window] [in a new window]   FIG. 1.   Estimated regression lines of log-based HIV load by CD4+ lymphocyte cell count for each population, excluding those whose estimated seroconversion date was within 6 months of the time that the viral load was measured. The cell-associated infectious HIV-1 load (A) was measured as log IUPM, and the plasma HIV RNA concentration (B) was measured as the log number of RNA copies per milliliter.

Slopes did not differ by group (P > 0.13 for each comparison) when relating HIV RNA load to CD4⁺ cell count (Fig. 1B). Neither female IDUs nor HM differed from male IDUs in CD4⁺-adjusted HIV RNA load as seen here by overlapping regression lines or in model 1 (Table 3). After pooling of the data for the groups, females had a lower level than males ( = 0.19; P = 0.037), and the risk group comparison was suggestive of higher levels in HM versus IDUs, but the difference was not significant ( = +0.11; P = 0.12).

Controlling for percent CD4⁺ cells rather than CD4⁺ cell count reduced the estimated difference in log viral load between HM and male IDUs in all comparisons to the point at which they no longer differed for infectious HIV load (P = 0.39 overall and P = 0.64 among n = 465) (Table 3). The difference estimates by gender within the IDUs, however, remained essentially unchanged (Table 3). Again, after pooling of the data for HM and male IDUs, women had almost a quarter log₁₀ decrease in infectious viral load compared to those for men ( = 0.24; P = 0.036); however, levels of HIV RNA were similar by gender ( = 0.11; P = 0.252).

To further explore why the difference estimates between HM and male IDUs in both viral load measures changed substantially after adjusting for percent CD4⁺ cells rather than CD4⁺ cell count, we evaluated the relationship between percent CD4⁺ cells and CD4⁺ cell count for each group. An increment of 100 CD4⁺ cells/µl corresponded to roughly an increment of 3.0 CD4⁺ cell percentage points, which was common for all three groups (P > 0.15 for each test). For a given CD4⁺ cell count, male IDUs and females had similar mean percent CD4⁺ cells (P > 0.5); however, HM had an estimated 4.5-percentage-point decrease in percent CD4⁺ cells compared to the percent CD4⁺ cells for IDUs combined (P < 0.01) (data not shown). Although the prevalence of smoking (87 versus 56%) and proportion of African Americans (97 versus 20%) were higher among IDUs than HM, respectively, these factors were considered and did not confound this difference (data not shown).

The estimated correlation between infectious viral load and HIV RNA load was 0.58 overall, 0.58 for male IDUs, 0.54 for female IDUs, and 0.59 for HM. Figure 2 presents the estimated regression lines relating both viral load measures for each group. The slopes did not differ by group (P > 0.35 for each test), with a common slope estimate of 0.67, implying a log₁₀ increase in HIV RNA load corresponds to roughly a two-thirds log₁₀ increase in infectious viral load. For a given HIV RNA load, HM had a higher infectious viral load ( = +0.16; P = 0.073) and female IDUs had a lower infectious viral load ( = 0.13; P = 0.233) compared to that for male IDUs. After combining data by gender among IDUs, the infectious viral load was 0.2 log₁₀ higher in HM than in IDUs (P = 0.011) for any given HIV RNA load.

View larger version (16K): [in this window] [in a new window]   FIG. 2.   Estimated regression lines of log-based infectious viral load, measured as log IUPM, by log-based HIV RNA load, measured as log number of RNA copies per milliliter, for each population, excluding those whose estimated seroconversion date was within 6 months of the time that the viral load was measured.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

A major finding of this study was that female IDUs tended to have the lowest levels of both cell-associated infectious HIV load and HIV RNA load compared to those for male IDUs and HM. Borderline significant gender differences among IDUs were observed only for infectious viral load. Differences between females and all males combined, however, were found to be statistically significant for both assays and after adjusting for either CD4⁺ cell count or percent CD4⁺ cells. Although these results are not conclusive in terms of differences by gender, they do tend to support earlier studies which indicated lower HIV RNA loads among women than men (3, 4, 9).

Katzenstein et al. (9) reported higher HIV RNA loads among homosexuals relative to those among nonhomosexuals, which consisted of mostly white males and females, after adjusting for CD4⁺ cell count. Our data also suggested that HM have higher viral loads than male IDUs on the basis of the results of both assays and when controlling for CD4⁺ cell count. However, the difference was statistically significant only for infectious HIV load. The differences were most pronounced when data for HM were compared to those for all nonhomosexuals. In contrast, when controlling for percent CD4⁺ cell viral loads in HM appeared to be similar to those in male IDUs. In addition, we showed that for a given CD4⁺ cell count, HM had a significant 4.5% decrease in percent CD4⁺ cells relative to the percent CD4⁺ cells among IDUs, suggesting that CD4⁺ cell count and percent CD4⁺ cells may not be used interchangeably to mark the stage of disease when comparisons are made across risk groups. Instead, percent CD4⁺ cells may be considered a more precise measure. The possibility of residual confounding from race or smoking status was considered, but these characteristics were not found to be confounders. Another explanation could be systematic laboratory variation between laboratory studies for the two cohorts, but this is unlikely since the lymphocyte counts and T-cell subset counts were performed in the same laboratory under the same protocol for both studies.

One potential limitation of the study results from the fact that 11% of the individuals did not have available specimens for quantification of plasma HIV RNA load. These subjects were more likely to be HM who were severely immunocompromised or who had AIDS, which would result in the observation of lower than expected viral loads among HM but not necessarily among male or female IDUs. Although the difference estimates by risk group may be conservative, attempts were made to control for differences by stage of HIV disease.

Two recent studies demonstrated a moderate correlation (r = 0.52 to 0.54) between cell-associated infectious HIV-1 load and plasma HIV-1 RNA levels as measured by PCR among predominantly white, homosexual, or heterosexual individuals (10, 18). These results are confirmed here with a combined population consisting of a larger percentage of African-American IDUs. Our results also suggest the correlation to be similar among HM, female IDUs, and male IDUs. In addition, the increase in infectious viral load given a log increase in HIV RNA load was estimated to be two-thirds of a log, which was common across all three groups.

Interestingly, we also found that when the copy numbers of HIV RNA in plasma were equal, HM tended to have significantly higher infectious viral loads than IDUs. This leads to two important questions that were beyond the scope of this study: Does infectious viral load have any prognostic ability independent of HIV RNA load? If so, do these higher infectious viral loads among HM relative to those among IDUs translate to faster HIV disease progression, despite similar levels of HIV RNA? Prior to the use of HIV load as a biomarker, two early Italian HIV-1 seroconverter studies (19, 20) found similar disease progression rates between HM and IDUs. Other studies (22, 24) have observed faster disease progression among the HM than among IDUs, although this was predominantly explained by high rates of Kaposi's sarcoma among HM. Two recent reports compared the usefulness of cell-associated infectious viral load as a predictor of HIV disease progression after adjusting for HIV RNA load within mostly white HM (6, 10). They found that when baseline virologic measures are available, infectious viral load is independently associated with disease progression, defined as a 50% decrease in the CD4⁺ cell count, AIDS, or death (10). If data from multiple time points are available, however, Fiscus et al. (6) show that the infectious viral load or changes in the infectious viral load are not predictive of disease progression independent of HIV RNA load, suggesting that the differences observed here may not necessarily relate to differential disease progression.

This was a cross-sectional analysis with mostly HIV-seroprevalent participants. Such an analysis has well-known limitations, including potential confounding due to different durations of infection between genders or risk groups. It is possible that if HM were infected earlier than IDUs in Baltimore and the relationship between viral load and CD4⁺ count changes over time, then our results could reflect an epidemiologic artifact rather than a basic biological difference. This could explain why the risk group differences in viral load were minimized when controlling for percent CD4⁺ cells. In contrast, the most consistent differences were observed by gender, even among IDUs only. Female IDUs did not appear to be more recently infected than male IDUs on the basis of the percentage of female IDUs with AIDS and percent recent HIV seroconverters (Table 1), suggesting real biological differences.

In summary, our data support differences in HIV load by gender, measured as cell-associated viral load or level of HIV RNA in plasma, and also suggest that the differences observed between risk groups may be driven predominantly by gender, because differences among males were only minimal. In addition, our data confirm the moderate correlation between cell-associated infectious HIV load and plasma HIV RNA copy numbers, which appears to be similar across both risk groups and genders.

The differences observed here, whether due to an epidemiologic artifact or to some biological mechanism, are consistent with earlier observations. While the observed associations await clarification through studies of longitudinal HIV load, the data do caution that clinical decisions related to the initiation of treatment with antiretroviral medications on the basis of a single viral load measurement need to consider the patient's characteristics. To date, clinical guidelines for the initiation of antiretroviral therapy have been generated from data derived from mostly white HM. Additional data on other groups could help to fine-tune guidelines.