The effect of sexually transmitted co-infections on HIV viral load amongst individuals on antiretroviral therapy: a systematic review and meta-analysis

A large body of evidence suggests that antiretroviral therapy (ART), particularly with newer treatment regimens, markedly reduces the risk of sexual transmission of HIV. Recent systematic reviews have estimated that ART causes a more than ten-fold reduction in the incidence rate within discordant couples, to less than 0.5 per 100 person-years [1‐6].

These sharp reductions have inspired the idea of antiretroviral treatment as prevention—aggressive programs to identify and treat HIV-positive individuals could substantially reduce HIV incidence at the population level, by reducing the infectiousness of HIV-infected individuals [5, 7]. However, increased infectiousness when treated individuals are co-infected with one or more other sexually transmitted infections (STIs) could potentially undercut the effectiveness of treatment as prevention programs. Concern with the effects of co-infection on HIV transmission is exemplified in the 2008 “Swiss Statement,” which argues that HIV sexual transmission risk is of no concern within stable discordant relationships in which: an HIV-positive partner is adhering to treatment under the care of a physician; the viral load has been suppressed for at least six months; and no other STIs are infecting the HIV-positive partner [8].

Although the biological mechanisms underlying this increased risk are not fully understood, many STIs have been associated with higher risks of both HIV acquisition and sexual transmission [9‐15]. Increased HIV transmission may be underpinned by higher HIV viral load resulting from larger concentration of HIV-infected immune cells in genital secretions induced by an inflammatory response and/or additional pathways caused by genital ulcers [16]. Similarly, inflammatory STIs, by recruiting immune cells, may provide additional targets for HIV virions, increasing HIV acquisition risk. Ulcerative STIs may present additional entry points for HIV infection [15].

Studies of HIV-STI interactions have been conducted mostly on individuals not receiving ART. Less is known about the impact of STI co-infections on HIV shedding from treated individuals. STI prevalence is high among HIV-infected individuals [17] and the proportion of these individuals on ART is quickly rising [18]. Thus, any potential increased HIV infectiousness due to STI co-infections among treated individuals could have important epidemiological consequences as treatment as prevention becomes more widespread.

Disentangling the many interacting factors at play is challenging: many STIs are suspected of affecting HIV shedding [9] and it remains unclear whether how these effects interact in people with more than one such infection; the viral load response to ART is regimen- and gender-specific [19]; numerous (not necessarily consistent) methods are used to sample and quantify HIV viral load [20, 21]; viral load measurements can vary between anatomical sites within an infected individual [22]; and HIV viral loads exhibit substantial temporal variation [23]. When considering transmission events in discordant couples, isolating the effect of STI co-infections is challenging because concomitance of STI infections in both partners (that could affect both HIV susceptibility and infectiousness) and HIV transmission are often not practical to ascertain.

Here, we conduct a systematic review and meta-analysis of the available evidence to assess whether STI co-infections affect the risk of HIV transmission from individuals on ART. We searched for studies that estimated transmission directly, and also for studies that measured viral load, which we intended to use as a proxy for transmission if the direct evidence was insufficient.

This systematic review and meta-analysis followed the guidelines from the PRISMA statement [24] (see Additional file 1). A protocol was prospectively registered in the PROSPERO database (see Additional file 2). Published peer-reviewed observational studies and randomized controlled trials were considered for inclusion. We included studies of sexually active HIV-infected participants on ART that were further classified into two subgroups: participants whose only known STI was HIV (the “mono-infected group”) and those with HIV and co-infected with another STI (the “co-infected group”). Studies were eligible for inclusion if they measured HIV viral loads among HIV-infected participants, or if they observed at least monthly HIV transmission events and STI infection status in discordant couples. An individual was considered co-infected only if the STI was laboratory confirmed. Individuals with ongoing treatment for the co-infecting STI were not included.

We searched for all relevant studies in Medline (Ovid), EMBASE (Ovid), PubMed, CINAHL and the Cochrane Library from inception to August 12^th, 2014. Subject headings and text words associated with the risk of HIV sexual transmission, ART and STIs were included in the search strategies. We included STIs commonly discussed in the context of HIV transmission: Chlamydia trachomatis, chancroid (H. ducreyi), any type of Human Papilloma Virus, Herpes Simplex Virus 2, Neisseria gonorrhoeae, syphilis (T. pallidum) and Trichomoniasis (T. vaginalis). We also included bacterial vaginosis and candidal vaginitis, although these are not known to result directly from sexual transmission, and urethritis, which can be associated with more than one STI. The search did not impose any language or geographical restrictions on studies. The STI positivity definitions are given in Additional file 3 and full search strategies in Additional file 4.

All retrieved abstracts were read by three authors (DC, CS and SH). Eligibility assessment was performed independently by two authors for each abstract, using pre-defined guidelines. Disagreement between authors was resolved by consensus after discussion. Data from eligible studies were extracted independently by two authors (DC and SH).

We sought to assess how two primary outcomes of interest, HIV viral loads and HIV transmission rates within discordant couples, varied between HIV-infected participants on ART with and without STI co-infections. We sought additional data to assess potential sources of bias within and between studies including STI diagnostic methods; anatomical sites sampled for HIV viral load measurements; HIV assays; interval between STI co-infection diagnosis and HIV viral load measurement; ART regimen, treatment length and adherence; study design; HIV-infected participant age, gender, and sexual orientation; and, for serodiscordant couple studies, concomitance of STI co-infection in HIV-uninfected partner and HIV genetic linkage following secondary partner seroconversion. Within studies, we excluded individuals not explicitly known to be on ART and, when this information was available, those who had been on ART for less than 30 days. We probed study quality by summarizing variables and methods (Tables 1 and 2), and with forest and funnel plots (Figs. 3 and 4).

Planned measures of effect were: the difference in log10 HIV viral load between the mono- and co-infected groups, for a given anatomical site and STI; and the relative HIV transmission rate between HIV-discordant couples with the infected partner belonging to the mono- versus co-infected group.

In the absence of qualifying studies that measured transmission risk directly, we conducted a meta-analysis of all available evidence of increased HIV viral load due to STI co-infection of individuals on ART, as a proxy for increased infection risk. Pooling information from all available studies, anatomical sites, and co-infections, we estimated that the average difference in viral load due to STI co-infection of individuals on ART was 0.11 log10 (95 % CI −0.62 to 0.83) greater than HIV mono-infected individuals.

Although our study provides some evidence for a small effect of STI co-infection on viral loads, we cannot rule out the possibility of no effect, or the possibility of a moderately large effect (the upper credible interval is 0.83 log10). Importantly, we are also not able to rule out the possibility that certain STIs (or certain combinations of STIs and anatomical site) have a much larger effect (see Fig. 3). Nonetheless, based on our analysis, we cautiously posit that ART manages—on average—to sustain its effectiveness at keeping HIV viral loads low during STI co-infection episodes, at the anatomical sites considered in this review (blood plasma, semen and cervicovaginal), and thus would be expected to maintain its effectiveness at preventing transmission.

Even with nearly 5000 data points used in this meta-analysis, realistic consideration of the variation between studies, STIs and anatomical sites reduces the statistical power considerably. There are other limitations that need to be highlighted.

Because eligible studies only estimated HIV viral loads, not transmission rates in discordant couples, our summary effect size is a proxy for HIV infectiousness and might not be an accurate representation of the actual sexual transmission risk. There were no eligible studies measuring rectal HIV viral load, so our estimate may not be applicable in assessing change in HIV infectiousness from receptive anal intercourse.

Not all studies tested for all STIs, hence we may have misclassified some co-infected individuals. This misclassification is likely, given the high prevalence rate of other STIs in the HIV-infected population [17]. Such misclassification could bias our estimates of the effect of STI on HIV viral load downwards.

Our study focused on whether STIs could interfere with viral suppression, and we therefore excluded HIV VL measurement done within 30 days of ART initiation. It seems likely that the effects of STIs on HIV viral load early in treatment are comparable to the effects before treatment begins and decline gradually through time, in an STI-specific fashion. We did not find any studies that could provide evidence bearing directly on the effect of STIs when ART has been initiated recently.

Adherence to ART was rarely reported or measured in the studies we used. Since the effect of STI co-infection on viral load would be stronger in people with poorly controlled viral loads, this effect is likely to bias our result. This bias will be exacerbated if co-infected individuals exhibit lower adherence (or could be reduced, or even reversed, if they are more likely to adhere, perhaps due to symptoms from the other STIs).

Most HIV viral load observations included in this meta-analysis were measured in blood plasma, not in genital secretions, which may limit the interpretation of our effect size to actual sexual transmission risk. Indeed, while plasma and genital HIV viral loads are correlated [44], evidences of HIV compartmentalization in some treated patients where viral loads as measured in genital secretions remain compatible with a non-negligible risk of transmission despite very low blood plasma viral loads [37, 45, 46].

We did not estimate gender-specific effects to avoid further complicating our model. It is possible that gender confounds our results to some extent. We note, however, that such confounding would be limited to blood measurements, since gender is implicitly accounted for in the other anatomical sites studied. Also, our estimate did not include a potential effect of menstrual cycle on genital HIV viral load, but we note that a recent study did not observe such effect [47].

If STIs were indeed an escape route for treatment as prevention, one could argue this should have shown up during large trials studying transmission rate among HIV discordant couples (for example [5]). But such trials do not provide complete reassurance, since trial participants are given STI monitoring and treatment not available to the general population.

The number of HIV-infected individuals receiving ART has dramatically increased during the last 10 years—from less than half a million in 2003 to about 13 million in 2013—and will increase further [18, 48]. The extended life expectancy associated with ART and the potentially higher exposure to STIs of the HIV-infected population may increase the prevalence of STI co-infections, particularly with public awareness of the decreased infectiousness of HIV while on ART and a potentially consequential decrease in condom usage. Hence, understanding the effects of such co-infections on HIV sexual transmission is an important public health issue.

Our study provides some insights into whether STI co-infections can undercut treatment as prevention efforts. Pooling available data, we estimate the degree to which STI co-infection may increase or decrease HIV shedding among treated individuals. We found a 95 % upper bound corresponding to a 0.83 log10 increase, which suggests that elevation of viral load by STI co-infections is unlikely to have a major impact on the ability of ART to reduce of HIV sexual transmission from patients on effective ART (as opposed to what could be observed in populations not on ART).

It is important to note that our results do not undercut the importance of control and treatment of STIs, not only for the well-being of infected individuals, but also for reducing HIV sexual transmission at the whole population level, which still has a majority of HIV infected individuals not on effective ART [18].

We did not have sufficient data to single out the effect of a specific STI on HIV viral load at a given anatomical site. Given the heterogeneous effect of STIs on HIV infectiousness, our analysis may have failed to identify epidemiologically relevant effects of particular STIs.

Hence, an important finding of our systematic review is that there is a paucity of available data with a sufficient level of detail to ascertain the effects of STI co-infection on the risk of HIV sexual transmission risk for individuals on ART. Future studies considering either transmission rate or HIV viral load may wish to consider the following suggestions: (i) given the possible high prevalence of co-infections among HIV-infected individuals, a broadest spectrum of relevant STIs should be tested; for transmission studies, testing should be performed frequently in both partners; (ii) HIV viral load should be measured in genital/rectal secretions, not blood plasma only; (iii) ART regimens and adherence should be reported, ideally at the patient level. We recognize that these suggestions may not always be practical, but when they can be followed they will help clarify the potentially important effects of STI-coinfection on the risk of HIV sexual transmission.

Our findings suggest that, on average, ART maintains its effectiveness at controlling HIV viral load during STI co-infections. However, with currently available data, we cannot rule out the possibility that certain STI co-infections have a larger effect. More high-quality studies specifically aimed at investigating the impact of STI co-infection on HIV sexual transmission from individuals on ART are needed.