Association Between Paired Plasma and Cervicovaginal Lavage Fluid HIV-1 RNA Levels During 36 Months

Abstract

Objective: 

To determine the patterns and predictors of genital tract HIV-1 RNA levels during a 36-month period.

Methods: 

HIV-1 RNA levels were measured blood in plasma and the genital tract (by cervicovaginal lavage [CVL]) at baseline before highly, active antiretroviral therapy, at 2 and 4 weeks and every 6 months. Viral loads were measured using nucleic acid sequence-based amplification assay with a lower limit of detection of 2.6 log₁₀ copies/mL.

Results: 

Ninety-seven women had a median of 30.4 months' follow-up, with 530 paired PVL and CVL specimens. The strongest predictor of CVL fluid HIV-1 RNA detection was PVL of more than 2.6 log₁₀ copies/mL, with an odds ratio of 13.7 (P < 0.0001). Each log₁₀ unit increase in PVL increased the odds of detecting HIV-1 RNA in CVL fluid by 2.6 folds (P = 0.0002). Cervicovaginal lavage fluid HIV-1 RNA exceeded PVL on 5% of visits. When patients achieved undetectable levels of HIV-1 RNA in both plasma and CVL fluid, rebound of HIV-1 RNA occurred in plasma first or concurrently with CVL fluid HIV-1 RNA.

Conclusions: 

Plasma viral load is the strongest predictor of CVL fluid HIV-1 RNA detection. Cervicovaginal lavage fluid HIV-1 RNA levels are generally lower than PVL. Plasma viral load is more likely to rebound first or at the same time as CVL fluid viral load.

Several studies have detected HIV-1 proviral DNA, cell-free RNA, and/or cell-associated RNA in the female genital tract with varied detection rates from 9% to 80%.^1-11 Further description of HIV-1 replication in both systemic and genital tract compartments is central to understanding transmission. Several authors have reported that plasma viral load (PVL) is a key determinant for HIV transmission.^1,2 It has also been suggested that the risk for sexual and mother-to-child HIV-1 transmission is related to direct contact with virus in the genital tract.^3,4 Plasma viral load has been shown to be a major determinant in predicting the risk for genital tract HIV-1 detection.^5-8 Most of these studies have been cross-sectional or have had brief follow-up periods (range, 1-12 months).^5-9 We determined the pattern of HIV-1 RNA in paired plasma and cervicovaginal lavage (CVL) fluid measurements during 36 months. We examined associations among CVL fluid HIV-1 RNA and PVL, antiretroviral therapy, genital tract infections, and the presence of semen in CVL. We assessed whether HIV-1 RNA rebounds first in plasma or in CVL fluid during failure or interruption of antiretroviral therapy after initial suppression.

METHODS

This study included HIV-1-infected women receiving care at the Miriam Hospital Immunology Center and Memorial Hospital (Providence, RI). Subjects were excluded if they were younger than 18 years, had recent pelvic surgery, were currently pregnant, or did not have a uterus or cervix. The study was approved by the institutional review boards of the Miriam Hospital and Memorial Hospital. Group A included women who were antiretroviral naive and beginning highly active antiretroviral therapy (HAART). Women in group B were nucleoside experienced and initiating HAART, whereas women in group C had no previous antiretroviral use and did not require therapy at enrollment. HAART was defined as any treatment regimen identified in standardized guidelines such as US Public Health Service or International AIDS Society-USA.

Baseline interviews included demographic characteristics and medical, sexual, and reproductive histories. Pelvic examination, tests for genital tract infections, and paired plasma and CVL samples were collected at each study visit. Visits were scheduled at baseline, 2 weeks, 1 month, and every 6 months thereafter. Visits were preferentially scheduled to occur at the midpoint of the menstrual cycle.

Nucleic acid sequence-based amplification (bioMérieux, Durham, NC) was used to measure HIV-1 RNA. All results are expressed as copies per milliliter with a lower limit of detection of 2.6 log₁₀ (400) copies/mL for both plasma and CVL. Semen in CVL fluid was assessed by the Abacus Diagnostic OneStep ABA card P30 test (Abacus Diagnostics, West Hills, CA).

The Spearman rank correlation coefficient was used to test associations between PVL and CVL fluid viral load. A single correlation coefficient was calculated at each follow-up visit, and the trend over time was examined. The association between a detectable CVL fluid viral load and a detectable PVL, as well as with other important clinical variables (age, race, genital tract infections, and semen), was measured through a logistic regression model. An interaction term between a detectable PVL and treatment group (A, B, or C) could not be supported with the available data; thus, a common odds ratio was estimated. To account for the correlation inherent in taking repeated measures for each subject over time, the logistic regression model was fit using generalized estimating equations. The model was fit using Proc Genmod in SAS v8.2 (SAS, Raleigh, NC). The resulting adjusted odds ratios with corresponding 95% confidence intervals were reported. The joint covariation of detectable CVL fluid viral load and detectable PVL was examined. The joint response at a particular visit, represented as time t, is denoted as [CVL_t, PVL_t]. We refer to the previous visit as t − 1 and denote the joint response at that time similarly. We considered possible transitions of the joint response of these variables from time t − 1 to the next available visit at time t.

RESULTS

Ninety-seven women had a median of 30.4 months' follow-up, with 530 paired PVL and CVL specimens. Thirty-six percent of women were white, 33% black, and 29% Latina, with a median age of 35 years. More than half (52%) of the participants reported a history of intravenous drug use, 67% reported a partner known to be HIV infected, and 36% reported exchanging sex for drugs or money. Most (76%) of the women were antiretroviral naive before enrollment. Twenty-four percent of women had previously received single- or dual-nucleoside therapy (group B). Forty-four percent did not require treatment at the beginning of the study (group C). During the course of the study, 9 women (21%) in group C started HAART, and these women were reclassified at the appropriate stage of analysis. At baseline, PVL ranged from 2.6 log₁₀ copies/mL or less to 6.0 log₁₀ copies/mL with a median of 3.8 log₁₀ copies/mL. Cervicovaginal lavage fluid viral load at baseline ranged from 2.6 log₁₀ copies/mL or less to 6.5 log₁₀ copies/mL, with more than half of the women having undetectable CVL fluid viral load. CD4 count at baseline ranged from 4 to 1517 cells/μL with a median of 446 cells/μL.

Figure 1 shows that, for each treatment group and for every study visit, a larger percentage of CVL samples had viral load at undetectable levels compared with plasma. A positive correlation between plasma and CVL fluid HIV-1 RNA was consistent and significant over time, as estimated by Spearman rank correlation coefficient (range, 0.40-0.44; P < 0.01), except for the final 2 visits (0.29, P = 0.08; 0.02, P = 0.89, respectively). HIV-1 RNA was undetectable in both plasma and CVL fluid in 41% of study visits (218/530). In 40% (214/530) of study visits, HIV-1 RNA was detected in plasma but not in CVL fluid. In 18% (93/530) of study visits, HIV-1 RNA was detected in both plasma and CVL fluid, with CVL fluid RNA greater than plasma in 25% (23/93). In 1% (5/530) of study visits, HIV-1 RNA was detected in CVL fluid but not in plasma, which occurred once among 5 different women. HIV-1 RNA was present in greater quantity in CVL fluid as compared with plasma on a total of 5% (28/530) of study visits. Fifty-five (57%) women had detectable HIV-1 RNA in CVL fluid. Of these, most (60%) contributed 1 positive sample, 38% had between 2 and 4 positive samples, and 1 woman (2%) had 5 positive samples.

FIGURE 1: Proposition of women with detectable virus in each compartment during every visit, stratified by treatment group, along with 95% exact binomial confidence intervals. Closed circles denote PVL, and open squares denote CVL.

In Table 1, the adjusted odds of having detectable HIV-1 RNA in CVL fluid were 13.7 times greater among participants with PVL more than 2.6 log₁₀ copies/mL (P < 0.0001). Among women with detectable PVL, the adjusted odds of having detectable CVL fluid HIV-1 RNA were 2.6 times greater for each log₁₀ unit increase in PVL (P = 0.0002). After controlling for potential confounding factors, there was no statistical evidence for an association between CVL fluid HIV-1 RNA shedding and trichomoniasis, Candida vaginitis, bacterial vaginosis (BV), or the presence of semen in lavage fluid samples. The observed trends and wide confidence intervals in our data suggest possible associations that we were unable to detect because of low power. Women were diagnosed with BV on 37% of study visits (194/517), 16% (86/527) with Candida vaginitis, and 6% (31/528) with trichomoniasis. Semen was identified in CVL fluid on 3% (15/526) of study visits.

TABLE 1: Adjusted Odds Ratios for the Presence of Detectable (>2.6 log₁₀ copies/mL) Cervicovaginal Viral Load*

We examined the variations in the proportions of detectable PVL and CVL samples by treatment group aggregating over all visits as shown in Table 2. To simplify the data presentation, we have combined the results of groups A and B, which were women initiating HAART. The probability of transitioning from no detectable virus in either compartment to having detectable PVL at the next visit was 18% and 32% for the combined group A/B and group C, respectively. By contrast, the probability of transitioning from no detectable virus in either compartment to having detectable viral load in the CVL fluid at the next visit was 3% and 13% for group A/B and group C, respectively. The probability of transitioning to a detectable viral load in the CVL fluid alone was unlikely from any previous state. Plasma viral load is more likely to rebound first or at least in tandem with CVL fluid viral load, occurring 100% of the time for group A/B and 95% of the time for group C, respectively.

TABLE 2: Variations in the Proportion of Detectable Plasma and Cervicovaginal Viral Load by Treatment Group

DISCUSSION

This longitudinal study confirms the findings of other studies that PVL is an important determinant of genital tract HIV-1 detection, and it addresses the durability of such an association. Although HIV-1 RNA shedding in the genital tract may occur in the absence of plasma viremia,^7,9,11 it appears that, in general, it is unlikely to find genital tract virus when PVL is undetectable. Several studies have also demonstrated a significant association between genital tract infections and genital tract HIV-1 shedding, whereas our current work did not.^6,7 A cohort with a higher prevalence of inflammatory genital tract infections might be expected to show a stronger association with HIV-1 genital tract shedding.

We did not detect a difference between groups A and B in PVL and CVL viral load patterns over time. Most women in group B were able to start HAART with a completely new regimen, which may explain the similar response to therapy as women in group A.

Our study was limited by the use of CVL fluid, a method resulting in dilution of mixed cervicovaginal secretions that may have affected our ability to detect virus in the genital tract more so than other newer techniques, such as cervical canal fluid absorbed by wicking methods. The CVL method has greater within- and among-subjects variability (log₁₀ SD HIV-1 RNA copies/mL) at 0.83 and 3.02, respectively, compared with endocervical wicking (Sno-Strip; Chauvin Pharmaceuticals Ltd, Ashton Rd, Romford, UK) at 0.55 and 1.66, respectively.¹¹ It should also be noted that genital tract HIV shedding is intermittent, and repeated measurements at shorter intervals may have shown different patterns. Newer ultrasensitive assays with lower limits of detection may have also altered the results. The use of whole CVL fluid did not allow for analysis by cell-associated versus cell-free RNA. Proviral DNA was not measured. Although no samples were collected during menses and all overtly blood-contaminated samples were discarded, occult blood was not quantified. Several studies, however, have shown that minimal blood contamination of genital samples does not significantly affect genital tract viral load levels.^6,12 We used wet mounts for assessment of lower genital infection, rather than more sensitive techniques, such as Trichomonas culture or polymerase chain reaction, or Gram stain for BV.

These findings have implications for HIV-1 transmission between sexual partners and from mothers to infants and substantiate previous research citing PVL as a key determinant of both sexual and vertical transmission of HIV-1. It is indeed plausible that, in general, women with undetectable PVL are less likely to express genital tract virus and thus are less likely to transmit HIV-1. Plasma viral load is a convenient clinical measure, and it appears to correlate strongly with genital tract HIV-1 RNA levels. This may also indicate that genital tract cell-free HIV-1 RNA is more dependent on changes in systemic HIV-1 RNA than on changes in the local genital environment as compared with cell-associated HIV-1 DNA that may be derived from local reservoirs. However, there are also numerous reports of discordance between blood plasma and genital tract viral load and also differences in diversity, divergence, and evolution of cell-free HIV-1.^7,9-11 The results of this study also confirm the occasional discordance between plasma and CVL fluid HIV-1 RNA, with CVL fluid RNA exceeding plasma on 5% of occasions. Longitudinal studies avoiding the limitations in this study may further increase our understanding of genital tract HIV-1 shedding.