Episodic Therapy for Genital Herpes in Sub-Saharan Africa: A Pooled Analysis from Three Randomized Controlled Trials

Background A randomized controlled trial in South Africa found a beneficial effect of acyclovir on genital ulcer healing, but no effect was seen in trials in Ghana, Central African Republic and Malawi. The aim of this paper is to assess whether the variation in impact of acyclovir on ulcer healing in these trials can be explained by differences in the characteristics of the study populations. Methodology/Principal Findings Pooled data were analysed to estimate the impact of acyclovir on the proportion of ulcers healed seven days after randomisation by HIV/CD4 status, ulcer aetiology, size and duration before presentation; and impact on lesional HIV-1. Risk ratios (RR) were estimated using Poisson regression with robust standard errors. Of 1478 patients with genital ulcer, most (63%) had herpetic ulcers (16% first episode HSV-2 ulcers), and a further 3% chancroid, 2% syphilis, 0.7% lymphogranuloma venereum and 31% undetermined aetiology. Over half (58%) of patients were HIV-1 seropositive. The median duration of symptoms before presentation was 6 days. Patients on acyclovir were more likely to have a healed ulcer on day 7 (63% vs 57%, RR = 1.08, 95% CI 0.98–1.18), shorter time to healing (p = 0.04) and less lesional HIV-1 RNA (p = 0.03). Small ulcers (<50 mm2), HSV-2 ulcers, first episode HSV-2 ulcers, and ulcers in HIV-1 seropositive individuals responded best but the better effectiveness in South Africa was not explained by differences in these factors. Conclusions/Significance There may be slight benefit in adding acyclovir to syndromic management in settings where most ulcers are genital herpes. The stronger effect among HIV-1 infected individuals suggests that acyclovir may be beneficial for GUD/HIV-1 co-infected patients. The high prevalence in this population highlights that genital ulceration in patients with unknown HIV status provides a potential entry point for provider-initiated HIV testing.


Introduction
Herpes simplex virus-2 (HSV-2) is the most common cause of genital ulcer disease (GUD) worldwide [1]. The natural course of genital herpes consists of subclinical genital viral shedding and clinical recurrences of blisters and ulcers which occur on the vulva, vagina or cervix in women, and on the glans, prepuce or shaft of the penis in men. The first episode of genital herpes is usually the most severe, lasting 2 to 4 weeks if untreated, and subsequent recurrences last 7 to 12 days without treatment [2]. A study conducted in the 1980s reported that oral acyclovir significantly reduced viral shedding, improved ulcer healing and decreased the incidence of new lesion formation among patients with recurrent genital herpes who received treatment within 48 hours of the onset of lesions [3]. A similar effect was seen on first episode genital herpes [4]. In 2003, the World Health Organization (WHO) published sexually transmitted infections (STI) treatment guidelines recommending addition of acyclovir to syndromic management of GUD in countries where the prevalence of HSV-2 as a GUD aetiology exceeds 30% [5]. However, a recent survey conducted in eight African countries found that acyclovir was rarely available in public health facilities, mainly due to cost and regulatory barriers [6].
The effectiveness of adding acyclovir to syndromic management for GUD in Africa was recently evaluated in three randomized controlled trials (RCTs). Two of the trials, one among women in Ghana and Central African Republic (CAR) and one among both men and women in Malawi, found no impact of standard dose acyclovir on ulcer healing [7][8]. However, a trial among South African men found that healing was significantly faster among those in the acyclovir arm (61% vs 42% healed at day 7; p = 0.003) [9] and acyclovir is now included in the guidelines for first-line management of GUD in South Africa [10].
In this paper, we analyze pooled data from the three RCTs to assess whether the variation in impact of acyclovir on ulcer healing can be explained by differences in the characteristics of the study populations in this large series of patients presenting with GUD in four distinct African settings.

Methods
Details of the three individual trials have been reported elsewhere [7][8][9] (Trials registration numbers: NCT00158483, ISRCTN32121857, NCT00164424). The trials had similar designs, with an a-priori plan to conduct pooled analyses. Briefly, in each trial, consenting consecutive patients with clinically confirmed GUD were enrolled and randomized to either acyclovir or placebo (400 mg three times daily for 5 days for Ghana/CAR and South Africa; 800 mg two times daily for 5 days in Malawi). Patients with very large ulcers (surface area $500 mm 2 ) and those with long ulcer duration (.21 days) were not included in the trial in Ghana/CAR and were treated immediately with acyclovir instead. Participants were asked to return for follow-up on day 2 and/or 4, day 7, day 10 (South Africa only), day 14 and day 28.
The primary outcomes in this analysis were the impact of acyclovir on the proportion of ulcers healed and lesional HIV-1 RNA (in HIV-1 seropositive patients) at 7 days post-randomization (defined as visits occurring on days 6-8 inclusive). The secondary outcome was time to ulcer healing over the 28 days of follow-up.
Ulcer healing was defined as a reduction in the size (in mm 2 ) of the largest ulcer at baseline by at least 90%. Further analyses used absolute ulcer size (,10 mm 2 ) as the outcome. Thresholds for detectable lesional HIV-1 RNA were 50 copies/mL for South Africa, 250 copies/mL for Ghana/CAR, and 400 copies/mL for Malawi. Ulcers which were not sampled because they were healed were assumed to have undetectable lesional HIV-1 RNA.
Statistical analyses were conducted in Stata 11.0 (StataCorp, Texas, USA). Poisson regression with robust standard errors was used to estimate risk ratios (RR) and 95% confidence intervals (CI) [11]. The risk difference and 95%CI for the primary analysis was also estimated, using binomial methods. Time-to-ulcer-healing was assessed graphically using Kaplan-Meier analysis and a log-rank test was used to obtain a p-value for a difference in time-to-healing by arm. The Kruskal-Wallis test was used to compare median durations between study sites. To investigate whether the differential effect of acyclovir in South Africa could be explained by factors associated with ulcer healing, we stratified analyses by South Africa versus non South Africa sites and tested for effectmodification by site. To allow for imbalances between arms, risk ratios of the impact of acyclovir were adjusted for factors potentially associated with ulcer healing (ulcer aetiology, size, duration of symptoms before enrolment and HIV status).

Characteristics of GUD patients
Of the 1516 GUD patients eligible for the three trials, 1478 were randomized (550 female and 928 male). The flow of participants through the trials is shown in Figure 1. Sociodemographic and behavioural characteristics of the participants are shown in Table 1 (by trial arm) and Table 2 (by site). There was good balance in characteristics by trial arm ( Table 1). The median age of participants was 28 years (interquartile range 24-34 years), and this was similar by study site and sex. Overall, almost half of men (47%) and one-third of women (33%) reported having had a previous GUD in the past year (Table 2). At the enrolment visit, the median number of ulcers per patient was two in each country, and women in Ghana and CAR tended to have fewer ulcers than patients in Malawi and South Africa. The size of the largest ulcer was measured for each patient, and varied considerably, with median size being smallest in Malawi (21 mm 2 and 28 mm 2 respectively in women and men), and largest in Ghana and CAR (147 mm 2 and 130 mm 2 , respectively), despite exclusion of ulcers $500 mm 2 in these sites. The median time from first symptoms to presentation was 6 days, and this was slightly shorter in South Africa and Ghana (5 days) than among women in CAR and women in Malawi (7 days; p = 0.0001). Overall, 28% of women and 13% of men were on antibiotics for the current ulcer, including 43% of women in CAR.
HIV-1 seroprevalence was high in all sites, ranging from 34% among women in Ghana to 77% among Malawian women (Table 3). Among HIV-1 infected persons, mean HIV-1 plasma viral load was 4.77 log 10 copies/mL (95% CI 4.71-4.84) overall, but varied significantly between site (p,0.0001), being lowest in Ghana (4.26 log 10 copies/mL), where the proportion of participants with CD4 count above 500 cells/mm 3 was also highest (38%). Overall, almost half of HIV-1 seropositive women (47%), and a third of HIV-1 seropositive men (34%) presented with CD4 counts below 200 cells/mm 3 . Despite this, the proportion of HIV-1 seropositive participants on antiretroviral therapy was low (9% in women, 3% in men). Overall, about half of the HIV-1 seropositive participants had detectable lesional HIV-1 RNA at enrollment, but this varied substantially by site (from 31% among women in Malawi to 80% in CAR; p,0.0001).

Impact of acyclovir on ulcer healing
Of the 1478 patients enrolled, 1258 (85%) were seen on day 7, of whom 1255 had a genital examination ( Figure 1). The proportion seen was similar by site (ranging from 82% among Malawian women to 89% in women in CAR) and by arm (85% in each arm).
Overall, there was a small impact of acyclovir on the proportion of ulcers healed on day 7 (63% vs. 57% in acyclovir and placebo arms respectively; RR = 1.10, 95% CI 1.00-1.20; risk difference = 5.7%, 95% CI 0.3%-11.1; Table 4). Adjustment for factors potentially associated with ulcer healing rate (GUD aetiology, size of ulcer at baseline, duration before presentation, and HIV/CD4 status) made little difference (adjusted RR (aRR) = 1.08, 95% CI 0.98-1.18). The impact on healing was greatest in South Africa (aRR = 1.21, 95% CI 1.05-1.40) compared with other sites (aRR = 0.98, 95% CI 0.87-1.10) with evidence of effect-measure modification of these adjusted risk ratios (p-interaction = 0.02). Overall, individuals randomized to the acyclovir arm had slightly faster healing (p = 0.04), principally due to the South African trial where time to healing was significantly improved (p = 0.007; Figure 3b). Acyclovir had little impact on time to healing in the other sites (p = 0.76; Figure 3a). Results were similar if ulcer size,10 mm 2 was used as the outcome (results not shown).
There was little difference in impact of acyclovir by duration of symptoms before presentation, but acyclovir had a better impact on healing of small ulcers than larger ulcers (p-interaction = 0.002; Table 4). Again, differences persisted between South African and the other sites (aRR for small ulcers in South Africa = 1.33, 95% CI 1.13-1.55; aRR for small ulcers in other sites = 1.10, 95% CI 0.95-1.26; p-interaction = 0.09).

Impact of acyclovir on lesional HIV-1 RNA
Of the 726 HIV-1 seropositive individuals who were seen on Day 7, data on lesional HIV-1 RNA was available for 667 (92%) ( Table 5). Overall, there was evidence of an impact of the intervention on detectable lesional HIV-1 RNA after adjusting for ulcer aetiology, size of ulcers, duration of symptoms and CD4 group (aRR = 0.73, 95%CI:0.55-0.98). Again, the impact was confined to South Africa (aRR = 0.58, 95% CI 0.39-0.84), with no evidence of an effect at other sites (aRR = 1.01, 95% CI 0.65-1.57; p-interaction = 0.05; results not shown).
Among the 466 HIV-1 seropositive individuals with HSV ulcers, there was strong evidence of an effect on the intervention on reducing prevalence of detectable lesional HIV-1 RNA (aRR = 0.66, 95% CI 0.47-0.92) and this was greater among the 64 participants with first episode HSV-2 ulcers (aRR = 0.20, 95% CI 0.03-1.11) compared to the effect among the 401 participants with recurrent HSV-2 ulcers (aRR = 0.74, 95% CI 0.52-1.02), although this difference was not significant (p-interaction = 0.13). The impact on both types of HSV-2 ulcers was stronger in South Africa, although the differences were not significant.
The impact on lesional HIV-1 RNA was also stronger in those presenting within a week of symptoms (RR = 0.54, 95% CI 0.33-0.90) and those with small ulcers (RR = 0.55, 95% CI 0.31-0.87). Overall, there was no evidence of an impact among the patients with CD4 counts above 500 cells/mm 3 or below 200 cells/mm 3 ,

Discussion
Herpes simplex virus type 2 is a recognised and important cofactor of the HIV epidemic [1,[12][13][14]. A number of RCTs have failed to demonstrate an impact of HSV suppressive treatment on acquisition [15][16] or transmission [17] of HIV, although acyclovir or valacyclovir have been shown to decrease plasma and genital HIV levels [18][19][20][21][22][23][24] and even to modestly reduce HIV disease progression [25]. A meta-analysis of 8 RCTs showed that high dose acyclovir ($3200 mg/day) significantly reduced mortality among HIV-infected individuals [26]. In this pooled analysis of episodic therapy trials, there was a small impact of acyclovir on the proportion of ulcers healed on day 7 and this was more substantial among patients with HSV ulcers, first episode HSV-2 ulcers, small ulcers, and HIV-1 seropositive individuals with CD4 counts between 200 and 500 cells/mm 2 . However, the impact was consistently greater in South Africa than in other sites, and this was not explained by different distributions of these clinical characteristics.
It is challenging to explain the greater impact of acyclovir on ulcer healing and lesional HIV in South Africa compared with other sites. The acyclovir regimen was the same in South Africa as in Ghana and Central African Republic (400 mg three times daily Defined as a . = 90% reduction in size of largest ulcer from enrolment. 2 Risk ratio for impact of acyclovir among people in that strata, adjusted for all other variables in the table (ulcer aetiology is categorised as non-HSV ulcer, HSV ulcer). 3 P-value for interaction. 4 Interaction between non-HSV and HSV ulcers. doi:10.1371/journal.pone.0022601.t004 for 5 days), and lower than in Malawi (800 mg twice daily for 5 days); the duration of GUD before treatment was marginally shorter in South Africa, but comparable to Ghana (median 5 days vs. 7 days elsewhere); and ulcer characteristics were not different in other respects (size and number) between South Africa and other sites. Reported adherence was high in all three trials (.90% of participants reporting having taken all their tablets by day 7), although no information was available on drug concentrations, nor on drug quality and storage. A limitation of the trials was that the first visit date at which all patients were recalled was one week after randomization, and we cannot evaluate impact on healing within the first week. A number of possible explanations have been put forward for the lack of efficacy of suppressive therapy on HIV acquisition and transmission, which include the inadequacy of current HSV treatment regimens to properly control HSV-2 replication itself [17]. It is possible that such a mechanism could affect ulcer healing as well. In addition, these trials have shown that African patients tend to come late for the management of their ulcers, which would compromise the efficacy of antiviral therapy. Clearly, earlier provision of more potent, and perhaps more prolonged, HSV-2 therapies are required to control HSV-2 replication and hasten healing, particularly among HIV-positive patients [27]. Earlier provision would be facilitated by counselling of patients with recurrent genital herpes to improve symptom recognition, as has been successful in the United States [28]. This pooled analysis provides descriptive data on a large series of genital ulcer patients from four countries in sub-Saharan Africa.
Almost two-thirds of ulcers were due to HSV-2, and of these, a quarter (225/918), were first-episode HSV-2 ulcers, representing also 16% of all GUD. Only 10% of all ulcers had a bacterial aetiology, alone or combined with HSV, although this varied by country. HIV-1 prevalence was high in these populations (58%), and a large proportion had CD4 counts below 200 cells/mm 3 , while few individuals were on antiretroviral therapy at the time of enrolment in 2003-2006.
The three trials were pragmatic trials designed to evaluate the effectiveness of adding acyclovir to current syndromic management of GUD, rather than the efficacy of acyclovir on treating herpetic ulcers per se. Although it is clear that there was a benefit in South Africa, the overall results suggest that there may be limited value in adding acyclovir to GUD syndromic management in some other African settings. Approximately seventeen GUD patients would need to be treated with acyclovir for one additional patient to be healed on day 7. There is likely to be some benefit in settings where a high proportion of ulcers are genital herpes, especially first episode HSV-2 ulcers which responded best to treatment, but this clinical and lab-based information is rarely available in resource-poor settings. The stronger effect among HIV-1 infected individuals than HIV uninfected individuals overall also suggests that acyclovir may be beneficial for GUD patients known to be HIV-infected at the time of presentation to clinical services. The high prevalence in this population highlights that genital ulceration in patient with unknown HIV status provides a potential entry point for provider-initiated HIV testing and care. It is not possible to diagnose HSV infection easily, quickly and cheaply, let alone primary HSV-2 ulcers, with current tools in resource-constrained clinical settings. Given the limited effectiveness of adding acyclovir to syndromic management overall, and the lack of clear reasons for why the regimen was more effective in one trial than the other two, further research is needed to explore better options for the management of GUD patients in sub-Saharan Africa. Adjusted for all other variables in the table.