Community-based trials of sexually transmitted disease treatment: repercussions for epidemiology and HIV prevention
Christopher P. Hudson1
This paper reviews the scientific basis for trials exploring the relation between sexually transmitted diseases (STDs) and human immunodeficiency virus (HIV) infection in Mwanza in the United Republic of Tanzania and Rakai and Masaka in the Republic of Uganda. The importance of a studys location and explanations for the divergent results of these trials are discussed. The modest effect on STDs seen in the trial of syndromic management in Mwanza, in contrast to the 38% reduction in the incidence of HIV, casts doubt on the underlying hypothesis that treating STDs alone slows the transmission of HIV-1. According to the PiotFransen model, the trial in Rakai, which offered treatment of STDs to all subjects irrespective of symptoms (mass treatment), should have been more effective both in reducing the prevalence of STDs and the incidence of HIV. However, the Rakai trial was stopped because there was no difference in the incidence of HIV between the intervention and control arms.
If Mwanza is seen as the trial that needs explaining, another paradigm becomes relevant. In rural East Africa, where all trials have been conducted, networks of concurrent sexual partnerships are a source of infection with both STDs and HIV. Because of their shorter latency periods, STDs may prompt attendance at a clinic before the early signs of HIV-1 infection appear. Part of the management of STDs is to recommend abstinence or the consistent use of condoms until treatment is completed. This recommendation may cover the earliest period of viraemia during primary HIV-1 infection. This paradigm appears to explain the results from Mwanza and Rakai, emphasizing behavioural aspects of syndromic management.
Keywords: sexually transmitted diseases, complications; HIV, transmission; HIV infections, therapy; comorbidity; randomized controlled trials; Republic of Uganda; United Republic of Tanzania.
Mots clés: maladies sexuellement transmissibles, complications; VIH, transmission; infections à VIH, thérapeutique; comorbidité ; essai clinique randomisé ; Ouganda; République-Unie de Tanzanie.
Palabras clave: enfermedades sexualmente transmisibles, complicaciones; infecciones por VIH, transmisión; infecciones por VIH, terapia; comorbilidad; ensayos controlados aleatorios; Uganda; República Unida de Tanzanía.
Introduction
In many ways, the public health response to AIDS has been more scientific than the responses to other diseases. The epidemic has coincided with moves towards practising evidence-based medicine and the now dominant use of randomized controlled trials in medical research. Thus, it would have been surprising if randomized trials of treatment for sexually transmitted diseases (STDs) had not been performed in the search for a better understanding of HIV transmission. There were specific reasons (13) why the two trial trials exploring the relation between STDs and HIV infection in Mwanza in the United Republic of Tanzania, and Rakai in the Republic of Uganda, were conducted (4, 5). However, there was no expectation that the results would diverge. An additional, similar trial will soon be completed in Masaka, Uganda, but this was planned long before publication of the divergent findings.
In Mwanza, the trial of STD treatment produced a 38% reduction in the incidence of HIV-1 infection despite there being only a modest effect on infection with STDs; in Rakai a trial of mass treatment of STDs produced no effect on the incidence of HIV-1 infection but there was a more marked, but still modest, effect on the incidence of STDs compared to Mwanza (3). Although more trials are planned or in progress, this paper argues that the Mwanza and Rakai trials provide enough information to revise the paradigm of the interaction between STDs and HIV. Because basic science and observational studies have already been comprehensively reviewed (3), this analysis focuses on the trials and discusses scientific method, the prevention of HIV transmission, and syndromic management.
Mwanza, Rakai and Masaka: three trials in East Africa
Between late 1989 and mid-1993 three randomized trials were designed to look at the interaction between STDs and HIV at the population level (2). The Mwanza trial began in 1992 and the other two taking place less than 300 km away (Fig. 1) started before the results of the Mwanza trial had been published.
The trial of improved syndromic management of STDs in Mwanza was designed to discover whether, at a population level, individuals infected with STDs are more likely to become infected with, or transmit, HIV (1) and to answer the practical question of how to improve the management of STDs in developing countries (2). The trial in Rakai was also a biomedical trial that is, there were no new behavioural interventions. The authors did not try to improve the management of STDs but instead sought to determine whether providing mass treatment in communities where there was a high incidence of HIV transmission, such as roadside settlements, has beneficial effect on rates of infection with STDs and the incidence of HIV infection. The third trial, in Masaka, compares the effect of health education alone with a Mwanza plus package (syndromic management of STDs plus health education) (2).
The hypothesis of the Mwanza trial was that inflammation associated with a symptomatic STD increased both infectiousness (person with STD and HIV positive but partner HIV negative) and susceptibility (person with STD and HIV negative but partner HIV positive). There was an impressive body of mostly epidemiological evidence at the time the Mwanza trial was designed (1). Recently, substantial basic science data have also accumulated (3). However, the hypothesis underlying the Rakai trial that asymptomatic carriage of a bacterial STD has a similar effect on HIV transmission is based on very limited evidence, at least in men. In women, data on the association between an asymptomatic STD and infectiousness is mostly qualitative (6, 7); additionally, data on viral load in genital secretions (8) are not as clear as data on men with a symptomatic STD (9). The scientific basis for the Masaka trial was the same as for Mwanza, with the added hypothesis that synergy exists when efforts to change sexual behaviour are conducted simultaneously with improved management of STDs.
Design and trial implementation
The designs of the three trials in East Africa are compared in Table 1 and baseline data (1013) are shown in Table 2. The trials in Mwanza and Masaka had input from researchers based at the London School of Hygiene and Tropical Medicine. It is therefore not surprising that these two trials are similar in design, and hopefully the results may be pooled for a meta-analysis. The Rakai and Masaka trials evolved independently despite being conducted in adjacent districts (Fig. 1), starting in the same year, and having head offices that are less than 100mapart (at the Uganda Virus Research Institute in Entebbe). Attempts to tabulate results of the trials in a comparative way have been hampered by the lack of common indicators other than HIV prevalence and incidence (8, 12). Even the approach to syphilis serology testing has been different: the Rakai trial used the toluidine red unheated serum test (TRUST test) for screening and then the Treponema pallidum haemagglutination assay (TPHA) for confirmation, whereas the Mwanza study used the rapid plasma reagin (RPR) and TPHA tests on all specimens.
Can results be compared?
Table 3 shows those results from Mwanza and Rakai that are directly comparable. The prevalence of active syphilis at the end of each study is shown: it is the most relevant public health measure. Table 2 includes both old cases and treated cases with low titres, thus giving a better picture of the history of syphilis in each population (14). The burden of syphilis was similar at the three trial sites at the start of the trials. Table 3 shows that the similarity in prevalence in Mwanza and Rakai persisted throughout the studies.
Table 2 and Table 3 highlight the importance of measuring several STDs to gain a picture of the stage of the HIV epidemic. Although rates of syphilis appear to have been similar in Mwanza and Rakai for some years, the relatively low prevalence of infection with HIV and herpes simplex virus type 2 (HSV-2) in Mwanza may mark the early stages of an HIV epidemic, and the high prevalence of these two viruses in Rakai may indicate a mature epidemic. The high prevalence of HSV-2 infection in Masaka, albeit measured by a now out-of-date methodology and only in two villages (11), indicates that Masaka and Rakai are at a similar stage of the epidemic. This is borne out by the similarity in mortality from HIVrelated causes (15, 16). In Mwanza the lower mortality among HIV-positive adults and lower awareness of AIDS deaths in the general population corroborate the assessment of the area as being in the early stages of an epidemic (17). However, the incidence of HIV and the prevalence of urethritis in the control arm of the study were half that found in Rakai, indicating that the potential for a large epidemic is limited.
Reconciling Rakai with Mwanza: conventional wisdom
Although comparing results from Mwanza and Rakai has been difficult, there is no mistaking that there was no difference in HIV incidence between the intervention and control communities in Rakai and there was about a 40% reduction in Mwanza. Most commentators on these studies believe that Mwanza is the gold standard and the Rakai results must be explained (3, 18). The issues raised in the most recent review (12) are discussed here.
Without doubt a major reason for the discrepant findings was the stage of the epidemic at the different study sites. The data from Rakai (19) confirm earlier predictions (20, 21) that the role of STDs in HIV transmission in the late stages of epidemics will be small compared to transmission in HIV-discordant couples (one partner HIV positive and the other HIV negative). Also associated with the stage of the HIV epidemic is the rise in HSV-2 infection (22) and, on the basis of serological testing, this common cause of genital ulcer disease would now appear to be a risk factor for HIV infection in Africa (23, 24). However, it could be that infection with HIV promotes transmission of HSV-2 (25) rather than the other way around. Although the viral load of HIV-1 rises when there are also ulcers caused by HSV-2 (26), those who have ulcers might abstain from sex once the lesions become painful sores. One study showed that there was no association between infection with HSV-2 and HIV, but an earlier survey at the same clinic found that there was an association (27). It may be that, like syphilis (22), the epidemiology of HSV-2 and HIV-1 diverge once a high HIV-1 prevalence/low HIV-1 incidence state has been reached at the population level. This is in keeping with findings from two villages in Masaka where there was no incidence of HIV-1 infection despite a high incidence of HSV-2 infection (11).
The lack of difference in treatment effects between men and women in the two trials (5, 28) is evidence that HSV-2 infection (note the large sex differences in serology in Table 2) and bacterial vaginosis did not act as confounders. The increased risk of HIV associated with bacterial vaginosis (if it is a genuine effect) is thought to operate by increasing susceptibility in women (29). However, among pregnant women in Rakai, in whom prevalence of bacterial vaginosis was 39% in the intervention group and 53% in controls, HIV-1 incidence was higher in women in the intervention group despite the significantly reduced incidence of bacterial vaginosis (5). The high rate of bacterial vaginosis in Mwanza (12) also militates against the idea that this condition contributes to the difference between the trials.
A major difference between the two trials was the way symptomatic STDs were managed. However, a high proportion of study participants in Rakai had symptomatic STDs (Table 3). If STDs have an important role in HIV transmission during chronic HIV infection (as opposed to acute primary HIV infection) some effect would have been expected in Rakai, provided reinfection rates were small. Wawer et al argued that reinfection is unlikely to be a sufficient reason for the negative findings of HIV incidence (5). Furthermore, reinfection is likely to have been a problem in Mwanza, given that only 2535% of index patients referred their partners for STD treatment (30, 31). Finally, while the small population attributable fraction (PAF) for STDs in HIV transmission in Rakai (32) is consistent with the negative result, the difference in the PAF in Mwanza between the intervention and control arms is only half the observed treatment effect (28). If the effect of the intervention in Mwanza had been mediated solely through treatment of STDs, the absolute difference in the PAF between treatment arms (23%) should have been similar to the effect on incidence of HIV-1.
The final explanation commonly given for the discrepant trial results is chance (8, 12). This relates to the imprecision of the estimates of intervention effect (the 95% confidence intervals), which in turn relate to sample size, study design, and imbalance of risk factors or prevalence of HIV and STDs in intervention-comparison pairs at baseline (matching of communities). The Rakai trial could have used six community pairs as in the other two trials but the awkward shape and topography of the region made this difficult. The reason for wide confidence intervals in the trials was the imbalance between the trial arms at baseline, which may have had a greater effect than originally thought (33). However, we should be wary of ascribing differences to chance just because the results of both trials conflict with the original hypotheses.
Mwanza and causality
The hypothesis of the Mwanza trial was that infection with an STD makes it easier for HIV to be transmitted. In the terminology of Rothman (34), STDs are intermediaries in the causal pathway of infection (Fig. 2). For individuals to become infected with HIV through sexual exposure they must have sexual contact with someone who is infected (the necessary cause). The risk of transmission is increased if a co-factor (the component cause) is present. However, there may be more than one component cause that, when combined with exposure, results in inevitable transmission (so that sufficient cause is met). This model suggests that an intervention aimed at an intermediary in the causal pathway must be more effective in eliminating that co-factor than in reducing the occurrence of the disease. In other words, Mwanza should have achieved a greater reduction in STD rates than in HIV incidence (35, 36).
Some may argue that after adjusting for baseline differences, a 49% reduction was achieved in the prevalence of symptomatic urethritis, and this is greater than the 38% reduction in HIV incidence. There are four responses to this. Firstly, the concept of causality implies large differences in the observed effect between the intermediary and the end result. Secondly, the reduction in urethritis was only a reduction relative to the control arm: the absolute prevalence went up when compared with baseline (37, 38). Thirdly, the basic scientific evidence for STD as a co-factor is for gonorrhoea (9), so symptomatic urethritis may not be the best measure. Some would add trichomoniasis to gonorrhoea on the list of causes of urethritis that affect HIV transmission (39).Fourthly, the epidemiological evidence has always been stronger for genital ulcer disease as a co-factor, and the effect on this was modest.
Mwanza: miracle or mirage?
Why did Mwanza work? There may be two answers: behaviour may have changed in relation to STD episodes and there may have been a decrease in the number of sexual partners. The Mwanza team may have missed a significant difference in sexual behaviour at follow-up by only interviewing a sample of 1 in 8 participants (40) and by only succeeding in interviewing 68% of those selected (28). Although the trial results were adjusted to account for differences in baseline variables (4), they could not be adjusted for important behavioural variables. It is possible that information campaigns in communities in the intervention arm resulted in women refusing new sexual partners. At baseline, 13% of women in the intervention group reported having had two or more sexual partners in the last year compared with 9% at the end of followup (4). In contrast, the proportion of women reporting two or more partners in the last year doubled in the control group from 7% at baseline to 14% at follow-up.
The rise in the number of partners may explain the apparent increase in the prevalence of STDs during the trial. Sexual behaviour in women, although under-reported, may be the most useful indicator of a populations risk of HIV (20, 41). It may have been the worsening economic circumstances in Mwanza (8) that put the population at increased risk. Alternatively, urban men may have begun seeking rural partners as a result of the visibility of AIDS in towns and the perception that rural areas had less of a problem (20). A study of factory workers in Mwanza town found that 79% knew someone infected with HIV (42); in the rural areas covered by the Mwanza trial only 7 of 73 deaths occurring among HIV infected adults were recognized by relatives as being associated with HIV (17).
While sample size calculations were based on an annual estimate of HIV incidence of 1% (40), this assumed that HIV prevalence was destined to rise. In fact, as elsewhere in rural Africa (43), prevalence might have stabilized at 4% with a low incidence but for the changes in behaviour suggested above. Between 1989 and 1991, antenatal seroprevalence remained stable in Mwanza town (44), and trends in rural areas generally follow those in towns after a 23 year delay (45). The Mwanza researchers could have tested the hypothesis that treatment effect depended on an increase in unsafe behaviour in the control group by seeing if the intervention pairs that showed the largest effect were those initiated last (the trial used a staggered introduction of pairs).
The Mwanza trial could have failed if all communities had lived in roadside settlements, as they did in Rakai. Fig. 3 shows the small impact of the intervention in the pair of roadside communities in Mwanza compared with the other community pairs (30). It seems that it is most difficult to intervene successfully in roadside communities, which are some of the most vulnerable and convenient communities in which to intervene (20). The benefit of intervening in roadside settlements may be felt chiefly in adjacent rural areas (20).
The PiotFransen model
The PiotFransen model (2) is shown in Fig. 4. This presents the burden of disease in a community, the proportion of infected individuals that are symptomatic, the proportion of these that seek care, and the proportion that receive appropriate care. If asymptomatic STDs are important in HIV transmission it is clear that the Mwanza trial would never have succeeded. However, if only symptomatic STDs have effects as co-factors, the main concern should be the small number of symptomatic patients who are correctly treated. In a post hoc assessment of the Mwanza intervention it was concluded that only 38% of participants in the intervention arm would have been cured (46).
The basis for the Rakai trial was the Piot Fransen model and the assumption that asymptomatic STDs have at least some effect in increasing HIV transmission. Rakai was a negative trial in terms of offering mass treatment as an intervention against STDs and in terms of preventing HIV transmission. With the exception of pregnant women (47), the effect on STD prevalence was marginal or, in the case of trichomoniasis, unimportant from a public health perspective (48). The widespread occurrence of asymptomatic urethral infections in men in Africa (39, 49) is in keeping with the view that repeated cycles of infection followed by imperfect treatment result in partial acquired immunity (50). The efficacy of mass treatment of STDs may depend on drug bioavailability in genital sanctuary sites (such as the prostate), and this may be impaired when there is minimal inflammation as a consequence of partial acquired immunity or the presence of scar tissue. Greater vascularity during pregnancy, with consequent drug penetration of genital mucosa, could perhaps explain the efficacy of the trial in pregnant women in terms of STDs.
Is Rakai the gold standard?
When a trial fails to show a significant result there are a number of possible explanations: the hypothesis could have been faulty; the location of the study could have been inappropriate; the study could have been poorly designed; or the study could have inadequate statistical power. In the Rakai study, the hypothesis would appear to be faulty, and cure rates for ciprofloxacin and azithromycin were assumed rather than proven. The decision to run the trial in an area where there is a mature epidemic that is, one in which the incidence is low relative to the prevalence has been heavily criticized. However, this scenario occurs in much of Africa, and thus this sort of trial could prove relevant to those making policy decisions, particularly in countries using azithromycin as preventive treatment for trachoma among the population.
The proximity of the Masaka trial and the possibility of contamination from their information, education, and communication programme and improved STD treatment arms, in addition to the presence of a Médecins du Monde programme in Kakuuto county (51) one of the areas included in the Rakai trial have never been discussed. However, as pointed out by the Rakai researchers (5), few participants seemed to have used appropriate STD treatment services, and there was no evidence of behaviour change. The latter negative finding may be due to the difficulty of capturing sexual networking using brief questionnaires; a comparison of two other studies in Rakai found a decrease in the extent of sexual networking over time (52, 53).
The design of the Rakai trial can be faulted for having intervention and control communities that were geographically too close together and for the fact that it did not include rural villages near to the study communities; these could have been a source of reinfection with STDs. The incidence of reinfection might also have been reduced if the interval between survey rounds had been shorter. However, as the Rakai researchers argued, reinfection is unlikely to have been a major factor (5). A better comparison of Mwanza, Masaka, and Rakai would have been possible if the Rakai trial had analysed data as a closed cohort study (excluding people seen in the second round who were not treated in the first). However, the negative outcome was not a case of missing a potentially significant difference (there is no problem with the study power): the trial was stopped early because incidence of HIV-1 was identical in both arms in both follow-up periods and the rate ratio was near 1 in all subgroup analyses (5). To be fair, the study could not have been performed to a higher standard.
Reconciling Mwanza with Rakai: a new hypothesis
Could the confusion over Mwanza and Rakai be a consequence of seeing Mwanza as the gold standard? New possibilities emerge if we regard Mwanza as the trial with the inherent contradiction (having more effect on HIV than STD) and Rakai as the trial with the expected result (no evidence for a role of asymptomatic STDs in increasing HIV transmission). There is evidence that roadside settlements in rural East Africa are the conduit by which both STD and HIV reach remote villages (20); there is also compelling evidence that sexual networking involving concurrent partnerships also acts as a conduit (5254). Men living in villages regularly visit roadside settlements and have sex with women there before returning to their steady partner or partners. If these sexual networks are the source of both STD and HIV then some people will contract both at the same time (Fig. 5). Due to their shorter latency periods, STDs may prompt clinic attendance before the viraemia of primary HIV-1 infection is apparent; this is believed to be the force driving the epidemic in Africa (20).
Part of the management of STDs is to recommend abstinence or the consistent use of condoms until treatment is completed. This recommendation may cover the earliest period of viraemia during primary HIV-1 infection. However, the infected individual may have transmitted an STD to one or more partners before seeking treatment, ensuring that STD prevalence remains high and HIV- 1 incidence is lower. This explanation of the Mwanza trial also shows why the groups who benefited most from the intervention were women aged 1524 and men aged 2534 (4); these are the groups most involved in sexual networking (20).
In both the Mwanza and Rakai trials there was a strong association between HIV transmission and genital ulcer disease among men who were HIV negative at baseline and who later seroconverted (28, 32). Ulcers could be a marker for primary HIV-1 infection (55). It is possible that pathogen-negative urethritis is also a symptom of primary HIV-1 infection (56), and in one African study of couples discordant for HIV infection 2 of 4 seroconversions in men were associated with urethritis that responded to monotherapy with co-trimoxazole, which is an inadequate therapy (57). Thus, associations between the incidence of HIV-1 and symptoms in men but not women (28) may reflect differences in manifestations of primary HIV-1 infection.
Abstinence during episodes of genital ulcer disease not only diminishes transmission during some cases of primary HIV-1 infection but also reduces transmission from individuals with advanced HIV-1 infection who may have pathogen-negative ulcers (58). Awareness of secondary syphilis as an STD may result in abstinence during episodes of rashes, which are known to be associated with transient, high titres of HIV-1 and disease progression (59). The role of brief episodes of viraemia in HIV-1 transmission should not be underestimated: in one small study the only marker of transmission risk in discordant couples was raised concentrations of suppressor T-cells (60). This finding was interpreted by the authors as indicating that there had been a recent episode of viraemia in the index case which had been controlled but not before transmission had taken place.
Rakai and the interaction between primary HIV-1 infection and STDs
In the Rakai trial mass treatment would have occurred at all possible stages of the scenario illustrated in Fig. 5. If a man received treatment while incubating both an STD and HIV, his partner would have been protected from contracting the STD but she could still have been infected with HIV-1. If he received treatment after developing and transmitting an STD then the scenario described in Fig. 5 would apply, except that the woman would have been treated and would have no increased susceptibility at the time of her partners viraemia. Thus, the incidence of HIV could remain high even though the incidence of STDs decreased.
Lessons for epidemiology
The first lesson from Mwanza and Rakai is to be sure of your hypothesis before conducting a trial. More basic scientific studies should have been conducted on asymptomatic STD infection and HIV viral load and the efficacy of single dose ciprofloxacin and azithromycin in curing chronic infection with gonorrhoea. The second lesson is to always collect process data so that a confusing or equivocal result can be interpreted. The Mwanza trial has been criticized for not proving directly that the average duration of STD episodes was reduced (35, 61).
The third, and most important, lesson is that the timing and location of a study are of supreme importance. Most scientists and funding agencies favour conducting studies where there is infrastructure and a body of data on which to base sample size calculations. It is ironic that Mwanza was the successful trial as there was inadequate infrastructure and comparatively little data before the trial. Insights into the local epidemic were based on annual antenatal surveys (44) and a cross-sectional community- based study (62, 63). Because only prevalence data were available, sample size calculations for incidence were based on assumptions. In contrast, there was a wealth of data from Rakai before the trial started (64, 65). With the benefit of hindsight, the Rakai team paid the price for choosing to stay where they were rather than move to a new area that was at an earlier stage of the epidemic.
Finally, and perhaps controversially, epidemiologists need to listen to the participants in their studies, have more faith in focus groups, and believe in behaviour change. The PAF of 10% calculated, at great expense, for the co-factor effect of STDs in the Rakai trial (32) is exactly the same as if it had been calculated on the basis of their pretrial cross-sectional study, which relied on self-reports (41). Furthermore, focus groups in early 1990 indicated that there had recently been a drastic change in sexual behaviour (65) when compared with previous sexual networking (54). This should have warned the Rakai researchers that the high incidence they observed in pre-trial data (64) was likely to be due to transmission occurring in couples discordant for HIV (21) rather than the co-factor effect of STDs.
Lessons for syndromic management
The rapid rise of tetracycline resistance in gonorrhoea cases from 35% to 96% in Mwanza should not be overlooked (30, 66). The monitoring of future interventions must incorporate susceptibility testing, and this must include the period after the trial has finished because standards of care may deteriorate. Follow-up studies in Rakai could define the risk to the future reproductive health of participants of untreated, asymptomatic STDs (the control arm of the trial). However, until and unless the follow-up of such patients confirms a substantial risk, the emphasis of syndromic management should be on providing counselling about abstinence and condoms and on contact tracing (22) rather than on obtaining the most expensive treatments and teaching staff how to use them.
Masaka: preparing for another negative result
Before they published their results, the Mwanza team outlined reasons why the trial might fail (40). Rather than waiting for a result and trying to explain it, this is the intellectually honest approach. This will be discussed using the sequence of issues raised by Hayes et al. (40), but the stage of the HIV epidemic and STD epidemic are added as key issues.
The slight upward trend in HIV-1 prevalence in Masaka (Fig. 1), the high seroprevalence of HSV-2 infection (Table 2), and the high HIV-related mortality (16) all indicated that there was a mature epidemic at the start of the intervention; there was, however, a lower prevalence of HIV-1 than in neighbouring Rakai. A natural history cohort outside the trial area has shown a significant decline in HIV-1 incidence (67), to well below the figure used in the calculations of sample size.
Contamination of the control group has occurred in Masaka. The original intention was to use a health intervention that was not associated with preventing HIV transmission (improving immunization coverage, malaria control, and sanitation) (8). However, there is now counselling available to those that request it and testing for HIV and condom social marketing in all arms of the trial (68). Additionally, the intervention has been introduced slowly, and the cure rate for STDs is only 70%. Only 70% of the target population was enrolled, so bias is likely. The slow pace of the intervention may relate to previous friction between researchers and study communities (69).
Conclusion
If the Masaka study yields a result that is not significant, the Mwanza study may remain the only trial of an intervention with a successful outcome. Yet the Mwanza trial may have been misinterpreted as confirming its hypothesis when it actually confirmed the behavioural component of syndromic management. A trial of improved counselling alone versus improved counselling and improved syndromic treatment is urgently required. Such a trial is needed in addition to ongoing studies.
Irrespective of the explanation for the success in Mwanza, the Rakai trial is bad news. In responding to the critics of the Rakai trial, Wawer et al. said: It is unlikely that a syndromic approach would have had substantial effects on HIV-1 in Rakai, where most HIV-1 transmission occurred independent of symptoms or laboratory diagnoses (61). Unfortunately, most of Africa has by now reached the same stage of the epidemic as Rakai had in 1994.
Laga argued in her 1995 commentary (70) on the original paper that reported the results of the Mwanza trial that rather than focusing our efforts on further quantifying the attributable risk of STD control, we should now put all our energies into determining how best to deliver STD care. The response to Rakai has been worthy of the title of Randy Shilts book And the band played on (71): the progress of science depends on revising hypotheses to fit research findings, rather than ignoring findings that do not fit hypotheses. The band must play on, but to a different tune: one that is behavioural not biomedical. 
Acknowledgements
Some of the points made and illustrations used first appeared in reviews by the author in the International Journal of STD and AIDS (20, 22). Permission from the editor, including permission to reproduce Fig. 3 and Fig. 5, is gratefully acknowledged. Fig. 4 was reproduced from ref. 2 with permission of the BMJ Publishing Group.
Résumé
Essais communautaires de traitement des maladies sexuellement transmissibles : répercussions sur lépidémiologie et la prévention du VIH
Cet article examine les bases scientifiques des essais visant à explorer la relation entre les maladies sexuellement transmisibles (MST) et linfection par le virus de limmunodéficience humaine (VIH), réalisés à Mwanza (République-Unie de Tanzanie) et à Rakai et Masaka (Ouganda). Limportance de la localisation de létude est soulignée et les explications des divergences entre les résultats des essais sont passées en revue. La modestie de leffet sur les MST dans lessai de prise en charge syndromique réalisé à Mwanza, par opposition à la diminution de 38% de lincidence du VIH, jette un doute sur lhypothèse selon laquelle le traitement des seules MST ralentirait la transmission du VIH-1. Suivant le modèle de Piot-Fransen, lessai réalisé a Rakai, et qui consistait en un traitement de masse des MST, aurait dû être plus efficace pour réduire à la fois la prévalence des MST et lincidence du VIH. Cet essai avait toutefois été arrêté du fait de labsence de différence dincidence du VIH entre les groupes dintervention et les groupes témoins.
Si lon considère lessai de Mwanza comme lessai aberrant nécessitant une explication, on se trouve face à un autre paradigme. Dans les régions rurales dAfrique orientale où ces essais ont été conduits, les réseaux de partenaires sexuels multiples constituent la source dinfection à la fois par les MST et par le VIH. Du fait de leur plus courte période de latence, les MST peuvent amener le sujet à se rendre dans un dispensaire avant que les premiers signes dune infection par le VIH-1 ne se manifestent. Dans le cadre de la prise en charge des MST, il est recommandé de sabstenir de relations sexuelles ou dutiliser systématiquement le préservatif jusquà la fin du traitement. Cette recommandation peut couvrir la période de virémie précoce lors dune primo-infection par le VIH-1. Ce paradigme semble pouvoir expliquer les résultats des deux essais, ce qui souligne limportance des aspects comportementaux dans la prise en charge syndromique de ces infections.
Resumen
Ensayos comunitarios del tratamiento de las enfermedades de transmisión sexual: repercusiones para la epidemiología y la prevención de la infección por el VIH
En el presente artículo se examina la base científica de los ensayos realizados para analizar la relación entre las enfermedades de transmisión sexual (ETS) y la infección por el virus de la inmunodeficiencia humana (VIH) en Mwanza (República Unida de Tanzanía), y Rakai y Masaka (República de Uganda). Se examina la importancia que reviste el lugar elegido para el estudio, así como las discrepancias existentes entre los resultados de esos ensayos. El discreto efecto en las ETS observado en el ensayo del tratamiento sindrómico llevado a cabo en Mwanza contrasta con el 38% de reducción de la incidencia del VIH y cuestiona la hipótesis subyacente de que el tratamiento de las ETS por sí solo disminuye la transmisión del VIH-1. Según el modelo de Piot-Fransen, el ensayo de Rakai, donde se ofreció masivamente tratamiento contra las ETS, debería haber sido más eficaz en lo que respecta a reducir tanto la prevalencia de las ETS como la incidencia del VIH. Sin embargo, el ensayo de Rakai se interrumpió porque no se detectó ninguna diferencia en la incidencia de infección por el VIH entre el grupo de intervención y el grupo testigo.
Si se considera que el ensayo anómalo es el de Mwanza, es posible concebir un mecanismo explicativo. En las zonas rurales de África oriental, donde se llevaron a cabo esos dos ensayos, las redes de parejas sexuales simultáneas son la fuente de contagio tanto de las ETS como del VIH. Dado su menor periodo de latencia, las ETS pueden llevar a la persona afectada a acudir al ambulatorio antes de la aparición de los primeros signos de infección por el VIH-1. Parte del manejo de las ETS consiste en recomendar la abstinencia o el uso sistemático del preservativo hasta el término del tratamiento, y esa recomendación puede abarcar el periodo inicial de viremia durante la primoinfección por el VIH-1. Este mecanismo parece poder explicar los resultados de los dos ensayos, y subraya los aspectos comportamentales del manejo sindrómico.
References
1. Mertens T, Hayes R, Smith P. Epidemiologic methods to study the interaction between HIV and STD. AIDS, 1990, 4: 5765.
2. Hayes R et al. Randomised trials of STD treatment for HIV prevention. Genitourinary Medicine, 1997, 73: 432443.
3. Fleming DT, Wasserheit J. From epidemiological synergy to public health policy and practice: contribution of other STD to sexual transmission of HIV. Sexually Transmitted Infections, 1999, 75: 317.
4. Grosskurth H et al. Impact of improved treatment of STD on HIV infection in rural Tanzania: randomised controlled trial. Lancet, 1995, 346: 530536.
5. Wawer M et al. Control of STD for AIDS prevention in Uganda: a randomised community trial. Lancet, 1999, 353: 525535.
6. Ghys PD et al. The associations between cervicovaginal HIV shedding, sexually transmitted diseases and immunosuppression in female sex workers in Abidjan, Côte dIvoire. AIDS, 1997, 11: F85F93.
7. Mostad SB et al. Hormonal contraception, vitamin A deficiency, and other risk factors for shedding of HIV-1 infected cells from the cervix and vagina. Lancet, 1997, 350: 922927.
8. UNAIDS/WHO. Consultation on STD interventions for preventing HIV: what is the evidence? Geneva, UNAIDS, 2000 (UNAIDS/ 00.06-WHO/HSI/2000.02).
9. Cohen MS et al. Reduction of concentration of HIV-1 in semen after treatment of urethritis: implications for prevention of sexual transmission of HIV-1. AIDSCAP Malawi Research Group. Lancet, 1997, 349: 18681873.
10. Kengeya-Kayondo JF et al. Risk perception and HIV-1 prevalence in 15,000 adults in rural south-west Uganda. AIDS, 1999, 13: 22952302.
11. Wagner HU et al. Seroprevalence and incidence of sexually transmitted deseases in rural Ugandan population. International Journal of STD and AIDS, 1994, 5: 332337.
12. Grosskurth H et al. Control of sexually transmitted diseases for HIV-1 prevention: understanding the implications of the Mwanza and Rakai trials. Lancet, 2000, 355: 19811987.
13. Mayaud P et al. Improved treatment services significantly reduce the prevalence of STD in rural Tanzania. AIDS, 1997, 11: 18731880.
14. Hudson CP. A population-based study of syphilis and sexually transmitted diseases in north-western Tanzania. Genitourinary Medicine, 1994, 70: 229.
15. Wawer MJ et al. Trends in HIV-1 prevalence may not reflect trends in incidence in mature epidemics: data from the Rakai population-based cohort, Uganda. AIDS, 1997, 11: 10231030.
16. Kamali A et al. Seven year trends in HIV-1 infection and sexual behaviour among adults in rural Uganda. AIDS, 2000, 14: 427434.
17. Todd J et al. HIV-associated adult mortality in a rural Tanzanian population. AIDS, 1997, 11: 801807.
18. Hitchcock P, Fransen L. Preventing HIV: lessons from Mwanza and Rakai. Lancet, 1999, 353: 513515.
19. Quinn TC et al. Viral load and heterosexual transmission of HIV-1. New England Journal of Medicine, 2000, 342: 921929.
20. Hudson CP. AIDS in rural Africa: a paradigm for HIV prevention. International Journal of STD and AIDS, 1996, 7: 236243.
21. Hudson CP. The importance of transmission of HIV-1 in discordant couples in the population dynamics of HIV-1 infection in rural Africa. International Journal of STD and AIDS, 1996, 7: 302304.
22. Hudson CP. Syndromic management for STD and HIV prevention: back to the drawing board. International Journal of STD and AIDS, 1999, 10: 423434.
23. Buve A et al. Risk factors for HIV in four African urban populations. XIII International AIDS Conference, Durban, South Africa, 914 July 2000 (Abstract WePpC1320).
24. Auvert B. HSV-2 is a major risk for HIV infection among young women in South Africa. XIII International AIDS Conference, Durban, South Africa, 914 July 2000 (Abstract LbPeD7127).
25. Mbopi Keou F-X et al. Genital herpes simplex virus type 2 shedding is increased in HIV-infected women. AIDS, 1999, 13: 536537.
26. Schacker T et al. Frequent recovery of HIV-1 from genital herpes lesions in HIV-1 infected men. Journal of the American Medical Association, 1998, 280: 6166.
27. Kharsany ABM, Mahabeer Y, Connolly C. Changing aetiology of GUD in STD attenders with rising HIV prevalence. XIII International AIDS Conference, Durban, South Africa, 914 July 2000 (Abstract ThOrC728).
28. Orroth K et al. Syndromic STD treatment reduces the proportion of incident HIV infections attributable to STD in rural Tanzania. AIDS, 2000, 14: 14291437.
29. SewankamboNet al. HIV-1 infection associated with abnormal vaginal flora morphology and bacterial vaginosis. Lancet, 1997, 350: 546550.
30. Mayaud P, Ka-Gina G, Grosskurth H. Effectiveness, impact and cost of syndromic management. International Journal of STD and AIDS, 1998, 9: S1114.
31. Attawell K, Grosskurth H. From knowledge to practice: STD control and HIV prevention. Brussels, Office for Official Publications of the European Communities, 1999.
32. Gray R et al. Relative risks and population attributable fraction of incident HIV associated with symptoms and treatable STD. AIDS, 1999, 13: 21132123.
33. Korenromp EL, De Vlas S, Habbema M. Effect of imbalances at baseline in community-randomized trials. XIII International AIDS Conference, Durban, South Africa, 914 July 2000 (Abstract WePpC1313).
34. Rothman KJ. Modern epidemiology. Toronto/Boston, Little, Brown and Company, 1986: 155175.
35. Rygnestad T et al. Impact of improved treatment of STD on HIV. Lancet, 1995, 346: 1157.
36. Whitaker L, Renton A. Impact of improved treatment of STD on HIV. Lancet, 1995, 346: 11581159.
37. Habbema JDF, de Vlas SJ. Impact of improved treatment of sexually transmitted disease on HIV infection. Lancet, 1995, 346: 11571158.
38. OReilly K et al. Impact of improved treatment of STD on HIV. Lancet, 1995, 346: 1158.
39. Jackson DJ et al. Urethral Trichomonas vaginalis infection and HIV transmission. Lancet, 1997, 350: 1076.
40. Hayes R et al. A community trial of the impact of improved STD treatment. AIDS, 1995, 9: 919926.
41. Hudson CP. Concurrent partnerships and the results of the Ugandan Rakai project. AIDS, 1993, 7: 286287
42. Ngweshemi JZL et al. Changes in male sexual behaviour in response to AIDS. AIDS, 1996, 10: 14151420.
43. Green SDR et al. Stable seroprevalence of HIV-1 in antenatal women in rural Bas-Zaire, 19881993. AIDS, 1994, 8: 397398.
44. Kigadye R-M et al. Sentinel surveillance for HIV-1 among pregnant women in a developing country: 3 years experience and comparison with a population serosurvey. AIDS, 1993, 7: 849855.
45. Hudson CP, Smith AN. High rate of p24 antigen in HIV antibody negative pregnant women in Kwazulu-Natal, South Africa. Presented at the XIIth World AIDS Conference, June 28July 3, Geneva, 1998.
46. Buve A et al. A closer look at Mwanza. XIIth World AIDS Conference, June 28July 3, Geneva, 1998 (Abstract 33234).
47. Wabwire-Mangen F et al. Randomized trial of STD control for maternal-infant health in Rakai, Uganda. XIII International AIDS Conference, Durban, South Africa, 914 July 2000 (Abstract MoOrC242).
48. Nicoll A et al. Preventing HIV1: lessons from Mwanza and Rakai. Lancet, 1999, 353: 1522.
49. Grosskurth H et al. Asymptomatic gonorrhoea and chlamydial infection in rural Tanzanian men. British Medical Journal, 1996, 312: 277279.
50. Brunham RC et al. Estimating the basic reproductive rates of Neisseria gonorrhoeae and Chlamydia trachomatis: the implications of acquired immunity. Sexually Transmitted Diseases, 1994, 21: 353356.
51. Kigongo D et al. Incredibly low knowledge among rural inaccessible communities in Rakai. XIII International AIDS Conference, Durban, South Africa, 914 July 2000 (Abstract LbPeD7127).
52. Morris M et al. Concurrent partnerships and HIV transmission in Uganda. XIth International AIDS Conference, July 7July 11, Vancouver, Canada, 1996 (Abstract TuD473).
53. Kelly R et al. Concurrent sexual partnerships and risk of prevalent and incident HIV. XIII International AIDS Conference, Durban, South Africa, 914 July 2000 (Abstract MoPpC1027).
54. Obbo C. HIV transmission through socialand geographical networks in Uganda. Social Science and Medicine, 1993, 36: 949955.
55. Kinloch-de Loes S et al. Symptomatic primary infection due to HIV- 1: review of 31 cases. Clinical Infectious Diseases, 1993, 17: 5965.
56. Bratt GA et al. Two cases of oral-genital HIV-1 transmission. International Journal of STD and AIDS, 1997, 8: 522525.
57. Kamenga M et al. Evidence of marked sexual behaviour change associated with low HIV-1 seroconversion in couples with discordant HIV serostatus. AIDS, 1991, 5: 6167.
58. Anderson J et al. Idiopathic genital ulcers in women infected with HIV. Journal of Acquired Immune Deficiency Syndromes, 1996, 13: 343347.
59. Fiscus SA et al. Transient high titers of HIV-1 in plasma and progression of disease. Journal of Acquired Immune Deficiency Syndromes, 1995, 9: 5157.
60. Seage GR et al. Increased suppressor T cells in probable transmitters of HIV. American Journal of Public Health, 1989, 79: 16381642.
61. Wawer M et al. Preventing HIV-1: lessons from Mwanza and Rakai. Lancet, 1999, 353: 15231524.
62. Mosha F et al. A population-based study of syphilis and sexually transmitted disease syndromes in north-western Tanzania. I. Prevalence and incidence. Genitourinary Medicine, 1993, 69: 415420.
63. Barongo LR et al. The epidemiology of HIV-1 infection in urban areas, roadside settlements and rural villages in Mwanza, Tanzania. AIDS, 1992, 6: 15211528.
64. Wawer MJ et al. Incidence of HIV infection a rural region of Uganda. British Medical Journal, 1994, 308: 171173.
65. Konde-Lule J, Musagara M, Musgrave S. Focus group interviews about AIDS in Rakai district of Uganda. Social Science and Medicine, 1993, 37: 679684.
66. West B et al. Antimicrobial susceptibility, auxotype and plasmid content of Neisseria gonorrhoeaein North Tanzania. Genitourinary Medicine, 1995, 71: 912.
67. Kamali A, Nabaitu J, Kinsman J. A randomized controlled community intervention to reduce HIV-1 infection in Uganda. XIII International AIDS Conference, Durban, South Africa, 914 July 2000 (Abstract WePpC1314).
68. Basajja V et al. A community-based condom social marketing strategy in rural Uganda. XIII International AIDS Conference, Durban, South Africa, 914 July 2000 (Abstract MoOrC244).
69. Seeley JA, Kengeya-Kayondo J, Mulder D. Communitybased HIV/AIDS research. Social Science and Medicine, 1992, 34: 10891095.
70. Laga M. STD control for HIV prevention. Lancet, 1995, 346: 518519.
71. Shilts R. And the band played on: politics, people, and the AIDS epidemic. New York, St Martins Press, 1987.
1 Honorary Lecturer, Department of Virology, University of Natal, Private Bag X7, Congella, Durban 4013, Republic of South Africa.
Ref. No. 99-0292