Sexual intercourse, mainly heterosexual, is the predominant route for HIV-1 transmission worldwide  and mother-to-child transmission has become an increasing public health problem in developing countries . In both cases, transmission is believed to occur through direct contact with the virus in the women's genital tract . As the inoculum size is an important determinant of virus transmission, the likelihood of perinatal and female-to-male transmission of HIV-1 is most likely associated with the amount of HIV in the cervix and vagina.
Studies regarding the presence, the amount and the determinants of HIV-1 RNA shedding in cervicovaginal secretions, in addition to the effects of antiretroviral therapy in its reduction, have thus important implications on the design of clinically based strategies for prevention of HIV transmission. Some reports suggest that factors altering the vaginal environment, such as certain genital infections [4–5], oral contraceptive use , and pregnancy  are associated with an increased HIV-1 shedding; moreover, virological and immunological co-factors, for example, plasma viraemia or the degree of immunodeficiency, seem to influence such shedding . On the contrary, there is a lack of information concerning the effects of antiretroviral therapy on the presence of HIV in cervicovaginal secretions. Recently, two publications analysed cervicovaginal HIV-1 shedding and its relationship to antiretroviral treatment in a significant number of patients; however, ultrasensitive techniques were not utilized for the quantification of plasma and cervicovaginal loads in one study  while a predominance of Latin-American and black women were enrolled among the patients in the other study , and thus it is not possible to draw definitive conclusions on this issue.
Herein, we report the results of a cross-sectional study in which the determinants of HIV-1 shedding in cervicovaginal secretions were analysed in a group of 122 infected Italian women with no clinical evidence of sexually transmitted diseases.
Materials and methods
Study population and eligibility criteria
The study population consisted of 122 HIV-1 infected women, consecutively enrolled from among the patients referring to the AIDS unit at the Clinic of Infectious Diseases, University of Bari, Italy, between December 1995 and June 2001. Eligibility criteria included HIV-1 seropositivity, age > 18 years and a normal Papanicolaou smear within the last 12 months. Criteria for exclusion included: (i) genital bleeding or vaginal discharge at the preliminary visit; (ii) active menstruation; (iii) pregnancy; (iv) use of oral contraceptives or intrauterine devices; (v) the presence of symptoms or signs compatible with genital infections at the preliminary visit or within the previous 4 weeks. Women were advised to refrain from vaginal intercourse and douching for the 48 h prior to the gynaecological visit, during which blood and cervicovaginal lavage (CVL) samples were obtained. Vaginal swabs for the detection of Candida and Mycoplasma infection by routine microbiological methods were also obtained.
With regards to antiretroviral therapy, the women were classified as follows: (i) under HAART treatment (when a treatment with two nucleoside reverse transcriptase inhibitors and one or more protease inhibitors or one non-nucleoside reverse transcriptase inhibitor was used); (ii) under mono-/dual-therapy; (iii) interrupted treatment (when the women had been treated with antiretrovirals, but, at the time of sampling, had interrupted the treatment for at least 3 months); (iv) naive (when the patient had never experienced antiretroviral treatment).
CVL were obtained by irrigating, after introduction of a speculum, the vaginal walls and the cervical areas with 3 ml of sterile 10 mM LiCl phosphate-buffered saline solution followed by aspiration, after 60 sec, from the posterior vaginal fornix . CVL and EDTA anticoagulated blood samples were processed within 1 h of collection; samples were centrifuged at 1200 rpm for 10 min and the cell-free CVL supernatants and plasma samples were stored frozen at –70°C until use. CVL were examined for the presence of contaminating semen by testing for the prostatic-specific antigen (PSA). In addition, in order to calculate the dilution factor introduced into the cervicovaginal secretions collected by the vaginal washing, Li concentration was evaluated before (Li1) and after (Li2) the vaginal washing, and the dilution factor  was approximated by the formula dx = (Li1)/(Li1)–(Li2).
HIV-1 RNA quantification
HIV-1 RNA quantification in CVL and plasma samples was performed with a commercially available technique (isothermal nucleic acid sequence-based amplification assay, NASBA; Organon Teknika Corp., Durham, North Carolina, USA), according to manufacturer's instructions. The lower detection limit of the test was 80 copies/ml.
Descriptive statistical techniques were initially used to provide a general profile of the study population in terms of CD4 count, plasma and vaginal HIV viral load (log10 transformation), exposure category, presence of an AIDS diagnosis and type of treatment. Scatter plot and Pearson's correlation coefficient were utilized to evidence an association between plasma and vaginal secretion HIV viral load (log10 transformation). Univariate and multivariate logistic models were performed to assess, respectively, crude odds ratios (OR) and adjusted odds ratios (AOR) for the presence of HIV-1 RNA in plasma and CVL. In the multivariate analysis, only variables significantly associated at the univariate level were considered. An undectectable level (below 80 copies/ml) of HIV-1-RNA was considered both for plasma and vaginal viraemia as the primary response.
In order to analyse the factors predicting the presence of HIV-1-RNA either in plasma or in the vagina, we stratified participants depending on the presence or absence of HIV-1 RNA in these two compartments into the following four groups: A, viral load undetectable both in plasma and vagina; B, viral load undetectable in vagina and presence of HIV-1-RNA in plasma; C, viral load undetectable in plasma and presence of HIV-1-RNA in vagina; D, presence of HIV-1-RNA in both plasma and vagina. The determinants associated with each group were then studied by using a polytomous logistic regression model and calculating the relative risk (RR) ratios using group D as reference.
All women enrolled in the study were white with a mean age of 33.5 years (range, 19–62 years) and a 51-month mean period of seropositivity (range, 9–120 months). The clinical stage of the disease, according to the Center for Disease Control Classification system was C (AIDS) in 43 (35.2%) and A (asymptomatic) in 79 (64.8%) cases. The risk category for HIV-1 infection was the use of injecting drugs in 45 cases (36.9%) and heterosexual intercourse with a seropositive partner in 77 cases (63.1%). At the time of enrolment, 75 women were not under antiretroviral treatment (54 were naive to antiretrovirals and 21 had suspended treatment because of therapy failure, side effects, intolerance or concurrent opportunistic infections), 28 were receiving highly active antiretroviral therapy (HAART), and 19 a non-HAART therapy.
Correlates of HIV-1 RNA presence in plasma
Of the 122 women enrolled in the study, 94 (77.0%) had detectable HIV-1 RNA in plasma samples; in particular, this was observed in 83.3% of women naive to antiretrovirals, in 100% of women interrupting treatment, in 73.7% of women subjected to non-HAART, and in 50.0% of women under HAART (Table 1). At both univariate and multivariate analysis, the presence of detectable virus in plasma was significantly associated with a decreased CD4 cell count (OR, 0.54 per each 100 cells count more; P = 0.001), injecting drug use (IDU) (OR, 6.27; P = 0.013), and receiving antiretroviral therapy when compared with antiretroviral-naive subjects (OR, 0.03 and 0.08, respectively for HAART treatment and mono/dual antiretroviral regimen; P ≤ 0.01), but not with the clinical stage of disease (AIDS versus non-AIDS), or presence of Candida or Mycoplasma in the vagina.
Correlates of the presence of HIV-1 RNA in cervicovaginal secretions
HIV-1 RNA was detected in 87/122 (71.3%) CVL samples; this occurred in 79.6% antiretroviral-naive women, in 100% of women interrupting treatment, in 78.9% of women under non-HAART regimens, and in 40% of women under HAART (Table 2). As for plasma viraemia, both at univariate and multivariate analysis, the presence of detectable HIV-1 RNA in CVL was significantly associated with a decreased CD4 lymphocytes (OR, 0.79 per each 100 cells count more; P = 0.029), and with HAART treatment when compared to antiretroviral-naive subjects (OR, 0.07; P = 0.001); no association was related to clinical stage of the disease (AIDS versus non-AIDS) or the presence of Candida or Mycoplasma in vagina. However, contrary to that observed in the plasma, a detectable viral load in the vagina was not correlated with IDU, or the receiving of mono- or dual-antiretroviral treatment.
Correlates of the combined presence/absence of HIV-1 RNA in plasma and cervicovaginal secretions
The scatter plot distribution of log10 HIV-1 RNA levels in plasma and CVL samples (Fig. 1a) demonstrates a significant linear correlation between the two variables [r2 = 0.56; 95% confidence interval (CI), 0.43–0.67; P < 0.001], thus indicating that the presence of HIV-1 RNA in plasma was associated with the presence of HIV-1 RNA in the vaginal secretions. Fig. 1a also shows 18 outlying data points; an analysis of the characteristics of these patients showed that among the 11 women with low vaginal viral load (log10 ≤ 2.0) and plasma viral load log10 3.5–7.0, seven were infected through sexual contacts and four by IDU, whereas all seven women with a low plasma (log10 ≤ 2.0) and vaginal viral load log10 2.5–7.0 were infected through sexual intercourse. When stratifying the correlation according to exposure category (Fig. 1b and c) a stronger correlation was observed in women who were infected through heterosexual intercourse (r2 = 0.61; 95% CI, 0.45–0.73) than by IDU (r2 = 0.44; 95% CI, 0.17–0.65), even if the two correlation indices were not statistically significant (P = 0.22).
When selecting the women with detectable HIV-1 RNA in both plasma and vaginal secretions as reference group, an adjusted analysis of the correlates of the combined presence of HIV-1 RNA in plasma and/or in cervicovaginal secretions (Table 3) showed that an increased CD4 cell count (RR ratio, 1.77 for each 100 cells more; P = 0.001) and HAART treatment (RR ratio, 61.43; P = 0.001) were significantly associated with non-detectable viral load both in plasma and in the vagina, whereas no association was found with other therapies or exposure categories (group A, 21 women). In the 14 women with detectable HIV-1 RNA in plasma but not in the CVL sample (group B), HAART treatment was the only significantly associated factor (RR ratio, 5.80; P = 0,02), whereas in the seven women with detectable HIV-1 RNA in CVL but not in plasma (group C), the increase in CD4 cells (RR ratio, 2.66 for each 100 cells more; P = 0.008) and non-HAART treatment were significantly correlated (RR ratio, 81.14; P = 0.02). Therefore, women under HAART treatment were more likely to reach undetectable viral levels in the vagina, even if HIV RNA was detected in the plasma, whereas women under non-HAART treatment were more likely to shed HIV in genital secretions even in the absence of plasma viraemia. Of interest, women in group C showed a very high mean CD4 cell count. Similar results were found when the presence/absence of HIV-1 RNA in plasma and in the vaginal secretion was considered using two independent logistic regression models (Tables 1 and 2). An increased CD4 cell count and HAART treatment were significantly associated with non-detectable viral loads both in plasma and in vagina. Non-HAART treatment was significantly associated with HIV-1 RNA absence in plasma viraemia but not in vaginal secretions.
In the present study, with the use of an ultrasensitive method allowing the detection of as few as 80 copies HIV RNA/ml, we analysed the determinants of HIV-1 RNA shedding in a group of HIV-infected women, homogeneous for ethnicity, in the absence of confounding variables, such as pregnancies, genital infections or use of hormonal or intrauterine contraceptives. Of the 122 women fulfilling the inclusion criteria and enrolled in the study, we demonstrated the presence of HIV RNA in CVL samples from as much as 71% of our study population. This result is in agreement with previous studies, in which cervicovaginal shedding of the virus was shown to be a common event during HIV infection (reviewed by Mostad ). Although lower rates of HIV recovery were reported by others, it is conceivable that methodological differences in CVL sampling and the use of more sensitive techniques to detect HIV RNA (such as that used in this study) may explain these differences.
We observed that the presence of a genital HIV shedding was positively correlated with plasma viraemia. This accounts for its suggested role as a marker of transmission for plasma viral load, both in mother-to-child  and heterosexual transmission  of the virus; in fact, one would expect that detectable plasma loads would be associated with higher genital viral loads, a factor influencing the rate of transmission. However, the results presented herein indicate that the virological scenario in the blood and genital tract differ in a certain number of patients, suggesting that these sites may be considered as different compartments.
In fact, 25% of the women with undetectable virus in plasma enrolled in the study showed a cervicovaginal shedding of HIV-1; of these, four were naive to antiretrovirals and three were under non-HAART treatment. Fourteen women presented the opposite situation demonstrating detectable plasma viral load in the absence of genital shedding; the majority of these subjects were naive to treatment or under HAART therapy.
Considered together, these indicate that in general, the viral load in genital secretions reflects the situation in the blood, so that viraemic patients more frequently shed the virus and, conversely, may be considered more at risk to transmit the infection both sexually and vertically; as a rule, an antiretroviral treatment capable of suppressing viral replication in the blood also suppresses HIV viral shedding at the genital site and thus antiretrovirals may be considered useful tools to reduce viral transmission.
However, from this study it can also be derived that important exceptions to the rule do exist and, namely, that in both treated and untreated women, plasma viral load may not be a reliable marker of transmission, since one-fourth of non-viremic patients in our study group were still shedding the virus in the genital tract. This finding confirms and extends previous observations from our group and that of others [10,19] and indicates that caution is required in judging infectivity of women on the basis of plasma viral load only, in both the sexual and mother-to-child transmission settings.
In addition, the present data indicate that different antiretroviral regimens may exert different effects on HIV shedding in cervicovaginal secretions. In fact, our results suggest that HAART treatment successfully reduces HIV viral load both in the plasma and vagina, although it appears to be more effective and rapid in the genital compartment. On the other hand, non-HAART treatment was found to be effective in reducing viral load in plasma but not at the cervicovaginal level. The interruption of antiviral therapy always resulted in detectable levels of HIV-1 RNA both in plasma and in vagina.
The role of high CD4 lymphocyte levels in reducing the viral load appeared to be more likely in the plasma than in the vagina. Therefore, a high CD4 cell count, even in the absence of plasma HIV-1 RNA (as shown in group C), does not necessarily imply the absence of HIV in the cervicovaginal secretions. A study conducted by Kovacs and colleagues  reported similar results; in particular, they observed a significant association between HIV vaginal shedding and both increased CD4 cell count and HAART treatment. However, when adjusting for plasma HIV RNA, these associations were no longer significant, probably due to an over-adjustment. In fact, plasma HIV RNA is associated both with certain exposure variables, such as CD4 cell count, antiretroviral treatment, and HIV shedding in CVL. Adjusting on plasma HIV RNA may lead to multi-collinearity, i.e., to a reduced statistical significance of the relative risk estimates which can lead to declaring detectable associations as insignificant.
We also found that women who acquired the infection through sexual contact showed a significantly lower plasma viral load when compared to those who inject drugs; available data does not allow speculation as to whether this is due to diverse biological factors related to mode of HIV acquisition or to a diverse compartmentalization of the virus. However, the finding that high CVL viral loads, in the absence of comparable values in plasma, were observed only in heterosexual partners is highly suggestive of a possible viral compartmentalization. Further studies to address this specific question are needed. An alternative explanation for the phenomenon might be a better adherence to antiviral therapy in heterosexual partners as compared with to injecting drug users.
There is increasing evidence that a local HIV replication occurs within the female genital tract and that this site may be considered as a distinct compartment for viral replication/evolution. Data presented here support this hypothesis and deserve consideration. Antiretroviral treatment was shown to dramatically reduce both heterosexual [15,16] and vertical viral transmission , probably via reduction of the viral load in the blood and genital secretions. It has been shown, however, that transmission may also occur even when antiretroviral agents are administered and a good response in the blood is achieved . This is most likely related to a lack of virological response at the mucosal level, that is, where transmission takes place. This possibility was documented in the present study and in previous papers from our group  and that of others ; according to our data, it appears to be more frequent in non-HAART treated women. In general, mono- and dual-drug regimens are no longer used for HIV infection, but infected pregnant women quite often undergo this kind of treatment. Plasma viraemia may thus fail to predict infectivity of genital secretions and therefore one should be cautious when using this parameter to judge the risk of transmission.
In our study, no correlation was observed between HIV shedding and the presence of cultivated vaginal Candida or Mycoplasma. Several authors reported increased levels of HIV-1 RNA in women suffering from genital infections [20,21] but recently their role has been revised . Moreover, one should consider that, for the purposes of the study, one eligibility criterion for patient enrolment was the absence of symptoms or physical signs of genital infections. It is thus conceivable that the mere presence of these pathogens in the genital tract may not cause an important inflammatory response capable of activating local viral replication.
We thank D. Potenza (Department of Infectious Diseases, Ospedale Perrino, Brindisi, Italy) for the enthusiastic collaboration in the initial phases of the project and for providing some of the samples enclosed in this study. We thank A. Lagioia for expert technical assistance, P. Butts for editing the manuscript and F. Iorio for secretarial assistance.
Sponsorship: Supported in part by grants from the Istituto Superiore di Sanità (Progetto AIDS).
1. Quinn TC. Global burden of the HIV pandemic.Lancet
2. Work group on mother-to-child transmission of HIV. Rates of mother-to-child transmission in Africa, America and Europe: results from 13 perinatal studies.J Acquir Immune Def Syndr Hum Retrovirol
3. Mostad SB, Kreiss JK. Shedding of HIV in the genital tract.AIDS
4. Wang CC, McClelland RR, Reilly MB, Overbaugh J, Emery SR, Mandaliya K, et al
. The effect of treatment of vaginal infections on shedding of Human Immunodeficiency Virus Type 1.J Infect Dis
5. Cu-Uvin S, Hogan JW, Caliendo AM, Harwell J, Mayer KH, Carpenter CC, et al
. Association between bacterial vaginosis and expression of Human Immunodeficiency Virus type 1 RNA female genital tract.Clin Infect Dis
6. Mostad SB, Overbaugh S, De Vange DM, Welch MJ, Chohan B, Mandaliya K, et al
. Hormonal contraception, vitamin A deficiency and other risk factors for shedding of HIV infected cells from the cervix and vagina.Lancet
7. Clemetson DB, Moss JB, Willerford DM, Hensel M, Emonyi W, Holmes KK, et al
. Detection of HIV DNA in cervical and vaginal secretions. Prevalence and correlates among women in Nairobi, Kenya.JAMA
8. Kovacs A, Chan LS, Chen ZC, Meyer WA 3rd, Muderspach L, Young M, et al
. HIV-1 RNA in plasma and genital secretions in women infected with HIV-1.J AIDS
9. Cu-Uvin S, Caliendo AM, Reinert S, Chang A, Juliano-Remollino C, Flanigan TP, et al
. Effect of highly active antiretroviral therapy on cervicovaginal HIV-1 RNA.AIDS
10. Kovacs A, Wasserman SS, Burns D, Wright DJ, Cohn J, Landay A, et al
. Determinants of HIV-1 shedding in the genital tract of women.Lancet
11. Mohamed AS, Becquart P, Hocini H, Metais P, Kazatchkine M, Belec L. Dilution assessment of cervicovaginal secretions collected by vaginal washing to evaluate mucosal shedding of free Human Immunodeficiency Virus.Clin Diagn Lab Immunol
12. Mostad SB. Prevalence and correlates of HIV type 1 shedding in the female genital tract.AIDS Res Hum Retroviruses
13. Leroy V, Montcho C, Manigart O, Van de Perre P, Dabis F, Msellati P, et al
. Maternal plasma viral load, zidovudine and mother-to-child transmission of HIV-1 in Africa DITRAME ANRS 049a trial.AIDS
14. Pedraza MA, Del Romero J, Roldan F, Garcia S, Acerbe MC, Noriega AR, et al
. Heterosexual transmission of HIV-1 is associated with high plasma viral load levels and a positive viral isolation of the infected partner.J AIDS
15. Musicco M, Lazzarin A, Nicolosi A, Gasparini M, Costigliola P, Arici C, et al
. Antiretroviral treatment of men infected with HIV-1 reduces the incidence of heterosexual transmission. Italian Study Group on HIV heterosexual transmission.Arch Intern Med
16. Nicolosi A, Musicco M, Saracco A, Lazzarin A. Risk factors for woman-to-man sexual transmission of HIV. Italian Study Group on HIV heterosexual transmission.J AIDS
17. Cooper ER, Charurat M, Mofenson L, Hanson IC, Pitt J, Diaz C, et al
. Combination antiretroviral strategies for the treatment of pregnant HIV-1 infected women and prevention of perinatal HIV-1 transmission.J AIDS
18. Ioannidis JP, Abrams EJ, Ammann A, Bulterys M, Goedert JJ, Gray L, et al
. Perinatal transmission of HIV-1 by pregnant women with RNA virus loads < 1000 copies/ml.J Infect Dis
19. Saracino A, Di Stefano M, Vimercati A, Greco P, Fiore JR, Monno L, et al
. Cervicovaginal HIV-1 shedding in pregnant women near delivery.Antiviral Ther
20. Lawn SD, Subbarao S, Wright TC Jr, Evans-Strickfaden T, Ellerbrock TV, Lennox JL, et al
. Correlation between human immunodeficiency virus type-1 RNA levels in the female genital tract and immune activation associated with ulceration of the cervix.J Infect Dis
21. Cohn MA, Frankel SS, Rugpao S, Young MA, Willett G, Tovanabutra S, et al
. Chronic inflammation with increased human immunodeficiency virus (HIV) RNA expression in the vaginal epithelium of HIV-infected Thai women.J Infect Dis
22. Rotchford K, Strum AW, Wilkinson D. Effect of coinfection with STD and of STD treatment on HIV shedding in genital secretions: systematic review and data synthesis.Sex Transm Dis