Genital Herpes Has Played a More Important Role than Any Other Sexually Transmitted Infection in Driving HIV Prevalence in Africa

Background Extensive evidence from observational studies suggests a role for genital herpes in the HIV epidemic. A number of herpes vaccines are under development and several trials of the efficacy of HSV-2 treatment with acyclovir in reducing HIV acquisition, transmission, and disease progression have just reported their results or will report their results in the next year. The potential impact of these interventions requires a quantitative assessment of the magnitude of the synergy between HIV and HSV-2 at the population level. Methods and Findings A deterministic compartmental model of HIV and HSV-2 dynamics and interactions was constructed. The nature of the epidemiologic synergy was explored qualitatively and quantitatively and compared to other sexually transmitted infections (STIs). The results suggest a more substantial role for HSV-2 in fueling HIV spread in sub-Saharan Africa than other STIs. We estimate that in settings of high HSV-2 prevalence, such as Kisumu, Kenya, more than a quarter of incident HIV infections may have been attributed directly to HSV-2. HSV-2 has also contributed considerably to the onward transmission of HIV by increasing the pool of HIV positive persons in the population and may explain one-third of the differential HIV prevalence among the cities of the Four City study. Conversely, we estimate that HIV had only a small net impact on HSV-2 prevalence. Conclusions HSV-2 role as a biological cofactor in HIV acquisition and transmission may have contributed substantially to HIV particularly by facilitating HIV spread among the low-risk population with stable long-term sexual partnerships. This finding suggests that prevention of HSV-2 infection through a prophylactic vaccine may be an effective intervention both in nascent epidemics with high HIV incidence in the high risk groups, and in established epidemics where a large portion of HIV transmission occurs in stable partnerships.

is the logarithmic (base 10) change in the viral load from an a viral load level to a b viral load level. That is, the 2.45 factor is the rate ratio increase in transmission probability with each one-log increment in viral load [6]. We use this relation in the estimation of HIV per-exposure transmission probability from dually infected persons during HSV-2 shedding, to HIV and HSV-2 susceptible persons (biological interaction parameters section below).

HIV prevalence levels:
As for the measured HIV prevalence level in Kisumu, Kenya, there is one notable population level survey, that of the four-city study, for the duration of June 1997 to March 1998 for the 15-49 years age group [7]. There are also antenatal clinic surveillance data provided by UNAIDS for the period of 1990 to 2002 [8]. The value of these data lies in providing HIV trends since they do not necessarily reflect the HIV population prevalence level during this time period [9]. We include these data points in our calculations (Figures 1-4) and use them to fit the trend while we use the only available population survey to fit the level in the year 1997-1998 for the sexually active population [7].

HSV-2 transmission parameters:
The estimates of HSV-2 transmission probability per coital act are in the range of 0.0005 to 0.022 depending on the nature of study. Prospective partner studies predict a value of 0.0005 [10] while time to HSV-2 studies estimate it at 0.022 [11]. Considering this variation, we derived three other independent but rough estimates for this probability.
The first is calculated using the cohort data of Corey et al. [12] by assuming uniform and average exposure across the partnerships over the follow-up period as can be seen in Table P2 S1. The probability per coital act is calculated from the partnership probability using the binomial model (  A third rough estimate is derived by using our population-level model for HSV-2 spread to fit observed HSV-2 prevalence levels in populations free from a generalized HIV epidemic. By varying the HSV-2 probability per coital act and comparing the predicted prevalences to those measured, we arrive also at a coarse estimate of 0.01 for the probability per sex act.
In view of these converging estimates, it is reasonable to assume that HSV-2 transmission probability per coital act is at 0.01 with a range of 0.0005 to 0.022. The exact value used in the model is chosen by fitting the observed HSV-2 prevalence levels.
Moreover, for lack of a biological mechanism, we assume that there is no HSV-2 transmission during the latent (no HSV-2 shedding) stage.
There are no measurements of the transmission probability per coital act during the primary HSV-2 infection. We assume that the transmission probability per sex act in this stage is equal to that in subsequent reactivations [15], but we assume that the primary infection lasts about twice as long as the reactivations [16]. Hence effectively the primary infection is twice as infectious as subsequent reactivations. Table P2 S3 displays our assumptions for the HSV-2 transmission probability per coital act.

HSV-2 shedding parameters:
We adopt the polymerase chain reaction (PCR) shedding data as the markers of HSV-2 infectiousness for consistency with the results that indicate ongoing transmission during periods of negative cell culture [17,18], as well as for consistency with model predictions of HSV-2 prevalence levels. A mathematical model that constrains transmission to only the duration when there is a positive cell culture, fails to predict the observed high HSV-2 prevalence levels. Cell culture data imply much lower prevalences than are actually observed.
The rate of HSV-2 sub-clinical shedding in subjects with no reported history of genital herpes is similar to that of subjects with a known history (3.0% versus 2.7% using cell culture) [19]. Indeed, the pattern, sites, and frequency of the sub-clinical reactivation of infection are similar across persons with or with no history of symptomatic herpes infection [19]. There is also evidence that HSV-2 is often transmitted during episodes of sub-clinical shedding [20,21]. Therefore, we assume that HSV-2 seropositive persons have the same infectiousness profile regardless of the presence of clinical manifestations.
We assume an average shedding frequency of 14% of the time in HIV seronegative patients based on the state of the art measurements of HSV-2 shedding that collected anogenital swabs for HSV-2 DNA PCR at four time periods per day for 60 consecutive days [22]. The primary HSV-2 infection results in substantially longer viral shedding than subsequent reactivations. Hence, we assume that this stage lasts for 20 days; about twice the duration of that of subsequent reactivations [16,23].

HSV-2 prevalence levels:
As for the measured HSV-2 prevalence level in Kisumu, Kenya, there is one notable population level survey, that of the four-city study, for the duration of June 1997 to March 1998 for the 15-49 years age group [24]. There are no available time series for HSV-2 prevalence for the period from 1980 up to present [25,26]. However, measurements from the early nineties in three communities that are in geographic proximity to Kisumu (Rakai [27] and Masaka [28], Uganda, and Mwanza, Tanzania [29]) as well as more recent measurements in Uganda [30], indicate similar HSV-2 prevalence levels as that of Kisumu in the late nineties. Moreover, retrospective analysis of sera from Zaire suggests that HSV-2 prevalence has grown steadily since the fifties but may have saturated at its current levels by the early eighties [31]. Therefore, we assume that HSV-2 prevalence has experienced only minor variability in Kisumu since 1980. The exact level is a prediction of the model fit.

Effect of dual infection on HSV-2 shedding frequency:
We assume that the HSV-2 shedding frequency in HIV infected subjects is 20% of the time for those in HIV acute and chronic stages (defined as CD4 count > 200 cells/μl) and 31% of the time in HIV advanced patients (defined as CD4 count ≤ 200 cells/μl).
These values are derived starting from the observed shedding in HIV seronegative persons [22], and then multiplying it by the observed fractional increase in HSV-2 shedding in HIV seropositives. HIV subjects shed HSV-2 40% more while in the chronic stage, and 120% more while in the advanced stage (compare the daily HSV-2 shedding of HIV seropositive persons in [32,33] to that of HIV seronegative persons in [12]). respectively) [32], these values may underestimate HSV-2 shedding since they are based on once-a-day sampling as opposed to frequent samplings per day [22]. Note that the increased shedding with HIV infection and declining CD4 cell count that we assume here is representative of the studies that measured HSV-2 shedding in HIV subjects [32,33,34,35,36,37,38]. These estimates also indicate no substantial differences in HSV-2 shedding in dually infected persons between women and men. Second, HSV family viruses has been observed to enhance HIV-1 transcription in vitro and in vivo [46,47,48,49,50], and HIV-1 RNA has been isolated with higher levels from herpetic lesions than from blood plasma [51]. Third, the clinical and sub-clinical herpetic lesions can disrupt the mucosal membranes thereby providing a portal for outgoing virons. There are however no concrete data to quantify the implications of these mechanisms on HIV-1 transmission probability per coital act. The "Partners in Prevention" study aims to shed light on this issue [52]. Earlier studies have suggested a role for dual infection in boosting HIV transmission with an overall relative risk ratio at or exceeding two for those who are dually infected with HIV and HSV-2 compared to those who are HSV-2 seronegative [53,54,55].

Dual infection of HIV and HSV
With these considerations in mind, we assume a three-fold increase in HIV transmission probability per sex act during HSV-2 shedding in dually infected persons based on the 0.5 log base 10 increase in HIV-1 plasma viral load with dual HIV and HSV-2 infection (that is a per-exposure cofactor effect due to enhanced transmission of 3 Trans PEC = ). For the sensitivity and uncertainty analysis (Protocol S3), we vary this enhancement from 2 to 5 as a plausible range for the variation of this effect.

Susceptibility to HSV-2 infection per stage of HIV infection:
There is limited evidence that indicates changes in the susceptibility to HSV-2 infection in HIV subjects [28,56]. It has been observed that previous HIV infection is a correlate of HSV-2 sero-conversion, but it is not clear whether this observation reflects a biological increase in susceptibility versus merely residual risk-behavior confounding or an increased risk of exposure to dually infected partners who have higher HSV-2 shedding rates.
Our model results indicate that even a factor of ten enhancement has little effect on the predicted HSV-2 and HIV prevalences (<2%). The reason is that HSV-2 is much more transmissible than HIV and has a much higher prevalence. The majority of HSV-2 infections occur before acquiring HIV. Therefore, in view of the lack of sensitivity to this effect and absence of concrete evidence, we assume that there is no increased susceptibility to HSV-2 infection in HIV seropositive subjects.

Susceptibility to HIV infection per stage of HSV-2 infection:
The nature of HSV-2 infection as a leading cause of clinical and sub-clinical genital ulceration and mucosal disruption, and the presence in herpetic lesions of CD4 lymphocytes which are the HIV target cells, suggest a role for HSV-2 infection in facilitating HIV acquisition [57,58]. This has been corroborated by numerous epidemiological studies that found a strong correlation between HSV-2 seropositivity and HIV acquisition even after controlling for sexual risk behavior. A recent meta-analysis including only longitudinal studies has determined that HSV-2 seropositive persons have an increased overall risk of HIV acquisition ( RR ) by a factor of 2.7 (95% confidence interval (CI),1.9-3.9) for males, and a factor of 3.1 (95% CI, 1.7-5.6) for females [59]. An earlier meta-analysis has arrived at a comparable value for the longitudinal studies of 2.1 (95% CI,1.4-3.2) for both males and females [60]. The analysis also found a risk estimate of 3.9 (95% CI, 3.1-5.1) in case-control and cross-sectional studies. In such studies, however, the temporal sequence of the two infections cannot be discerned.
The above estimates do not specify the increased risk per HSV-2 stage (primary infection and reactivation versus latent). We assume that in the latent stage there is no increase in susceptibility in view of the lack of a plausible biological mechanism.
Assuming that HSV-2 seropositive persons shed the virus at a frequency of 14% of the time [22], the susceptibility enhancement during HSV-2 shedding has the value of 8.9 if 2.1 RR = , and 14.6 if 2.9 RR = (average value for RR over males and females in [59]).
Noteworthy is that the increased risk of HIV acquisition during shedding is an order of magnitude larger in value than the overall relative risk RR .
Biological per-exposure cofactor effect due to enhanced susceptibility to HIV acquisition: We calculated the per-exposure cofactor effect ( PEC ) using the methodology of Hayes et al. [61] which links the measured overall relative risk ratio RR to the transmission probability per partnership, and subsequently to the transmission probability per coital act using the binomial (Bernoulli) model [2]. This is done by expressing the overall relative risk of HIV acquisition in HSV-2 seropositives relative to those HSV-2 seronegative ( RR ) in terms of HIV transmission probability per partnership as where 0 z is the HIV transmission probability per partnership if the HIV susceptible partner is HSV-2 seronegative while 1 z is that if the HIV susceptible partner is HSV-2 seropositive ( ) Here we assume an average HIV transmission probability per coital act of 0.0015 p = Here, The seminal Rakai data provides an alternative, and independent, avenue to derive the Acq PEC . It was found that the average HIV transmission probability per coital act was five fold higher with HSV-2 seropositivity (0.002 vs. 0.0004, 0.01 P = ) [33,54].
Assuming that in the partnerships where the susceptible partner was HSV-2 seropositive, the source partner was also HSV-2 seropositive, then we can derive an estimate of the Acq PEC during HSV-2 shedding. The transmission probability per partnership in a partnership between a dually infected person and an HSV-2 seropositive, but HIV seronegative, person is given by Here, 0 0.0004 p = is the average HIV transmission probability per coital act in absence of HSV-2 per exposure cofactor [33,54], and 40 P τ = months is the duration of follow-up in the Rakai study [1]. The 3 z probability can be also expressed as  [59]. Lastly, for the sensitivity and uncertainty analysis (Protocol S3) we assume a plausibility range of 3 to 10 for the Acq PEC to span the plausibility range for this parameter.

Dual infection of HIV and HSV-2 impact on HSV-2 transmission probability per coital act:
Due to lack of data and absence of manifest biological mechanism, we assume that coinfection with HIV does not increase HSV-2 transmission probability per coital act.

Effect of dual infection on the natural history of HIV infection:
There is lack of evidence to show that dually infected subjects progress faster to AIDS. However, treatment with acyclovir does not seem to prolong significantly survival to AIDS among dually infected individuals [66]. Therefore, we assume for simplicity that dual infection has no effect on HIV disease progression.

Behavioral input of the model
The parameters that describe the key behavioral characteristics in Kisumu, Kenya are based on the comprehensive measurements of the four-city study [7,67,68,69,70]. A summary of these measures can be found in Table S3 in the Supporting Online Material of Abu-Raddad et al. [3]. We use these measures to inform the sexual behavior parameters of our model. Note that it is very complex to quantify sexual risk behavior considering the diversity of sexual behavior measures and the multitude of facets of human sexuality.
We divide the population into two sex-risk groups. The fraction of people who are in the high risk ( high f ) versus the low-risk population is taken as the average of the following quantities: 1) proportion of men (33.5%) and of women (5.9%) who reported more than one partner (spousal or non-spousal but excluding sex workers) in the past 12 months among the sexually active population [69], 2) the proportion of men (19.5%) and of women (4.1%) with more than one non-spousal partnership (excluding contact with sex workers) in the past 12 months [68], 3) the proportion of men (3%) who had contact with female sex workers in the past 12 months [70], and 4) the number of female sex workers per man aged 15-49 years (1.95%) [68,70]. Hence we arrive at 11.3% as a representative value for the fraction of the core group in the population for both males and females. This estimate is reasonable considering that the high risk group is a minority in the population.
We assume for simplicity that the new sexual partner acquisition rate is independent of HIV or HSV-2 infection status but depends only on the risk group status (high risk group versus the low-risk population). However, the frequency of coital acts does vary depending on HIV stage of progression as measured by Wawer et al. [1] and tabulated in Table P2 Note that the effective rate of partner change does not merely reflect the actual rate at which individuals change their partners but also represents other behavioral mechanisms that effectively enhance this rate such as concurrency and topology of sexual networks [71,72,73], and variability in risk behavior [74] (see brief discussion in Protocol S1).
There are no measurements of the assortativeness in the mixing between the risk groups in Kisumu. However, the behavior measures in [68,69,70], such as the mixing with female sex workers, suggest a limited degree of assortativeness relative to proportionate mixing. Therefore, motivated by the model fit, we adopt a value of 0.2 e = for the assortativeness parameter.
The duration of sex partnerships depends on the sexual-risk group. We assume the duration of partnerships between members of the high risk group to last for 1 month.
Meanwhile, partnerships in the low-risk population last for 36 months, and partnerships between a member of the high risk group and a partner in the low-risk population last for 6 months. These assumptions reflect the long duration of spousal, and to some extent non-spousal partnerships excluding contacts with sex workers (median non-spousal is 11 months in Kisumu) [68], and the variable but generally short-duration partnerships with sex workers [70].
The duration of the sexual lifespan ( Τ ) is set at 35 years to conform with the 15-49 years age groups that is typically used to define the sexually active population by the WHO as well as many HIV studies [75]. Hence, the removal rate from the sexually active population is 1

Summary of the biological, behavioral and demographic parameters
The parameters that describe HIV transmission and progression along with the perexposure cofactors are tabulated in Table P2   2.0 years [1] HIV progression rates between stages: From acute to chronic stage ( persons in terms of the acquisition per-exposure cofactor ( Acq PEC ) rather than directly using the relative risk ratio ( RR ) (further discussion above on the per-exposure cofactor derivations and one comment regarding this parameterization in Protocol S1).
9.0 (derived based on meta-analysis in [59] for Meanwhile the duration of reactivation within the cycle is provided with Please note how the durations of latency and reactivation depend on HIV status and stage. Although there are substantial variations in the pattern (and frequency) of reactivations [16,22], we found in our model that the critical parameter here is the shedding frequency irrespective of whether the pattern is that of short but frequent reactivations or long but less frequent ones. This has been found by keeping the shedding frequency fixed, but varying the pattern of shedding. The model predictions were invariable.
The rates of HSV-2 progression ( Q β π ) from one stage ( β ) to the next for any population variable Q β are derived from the durations of each stage 1 Table P2 S3 Summary of the values of HSV-2 transmission and shedding parameters in our model.    Table P2 S4 shows a summary of our choices for the behavioral and demographic parameters in the model. We assume an initial host population size of 200,000 in Kisumu, Kenya as the representative average adult population from 1980 to present in absence of HIV mortality [70,76,77,78].