Microbicides could increase HIV infections in certain circumstances, models predict

Mathematical models predicting the effect of microbicides show that their effect on HIV incidence and prevalence could vary enormously, according to local conditions. In some cases their adoption could cause an increase in HIV infections.

Marie-Claude Boily of Imperial College, London, speaking at a symposium on modelling at the 2010 International Microbicides Conference in Pittsburgh, said that models predict some paradoxical effects.

A microbicide may have much more effect in a low-prevalence area than a high-prevalence one; it might reduce HIV infections in people who don’t often use condoms while considerably increasing infections in people who use them consistently; and a rectal microbicide, used only for the occasional episode of anal sex, could in some heterosexual populations have more effect than a vaginal microbicide used frequently.

Two things stood out from the models. If microbicide adoption results in anything more than a slight drop in condom use, the result could be disastrous for HIV prevention; and even a small amount of anal sex in a heterosexual population may slash the effectiveness of a vaginal microbicide.

Boily said that modellers put a number of variables into their models:

• The general properties of microbicides (safe, acceptable, efficacious, affordable);
• their specific properties (vaginal and/or rectal, whether they act on STIs as well as HIV;
• whether they are bidirectional, i.e. protect the insertive partner too);
• the properties of the community that uses them (availability, adherence, anal sex levels); and
• potential negative changes after introduction (drug resistance, condom substitution).

There had been eleven published modelling studies of the effect of vaginal microbicides in women, including ones presented at this conference.

A summary of four of them show that if modest efficacy and realistic usage levels are fed into models, the result in the absence of other confounding factors is a modest decrease in infections. Another complicating factor is that this reduction is measured in different ways.

For instance:

• A 2005 model (Smith) of microbicide use amongst female sex of a microbicide of 50% efficacy, used 43% of the time, and with no condom substitution, resulted in a reduction in the risk of infection of 17%.
• A 2006 model (Vickerman) of a microbicide with an efficacy of 40% both against HIV and against STI infections, used 37.5% of the time, and with 5% of the population ceasing to use condoms, produced a 14% decline in HIV incidence over four years.
• A model presented at this conference (Foss) of a rectal microbicide with 35% efficacy and 30% uptake with no condom substitution found that it would prevent 14% of infections over a five-year period.
• A model used to generate data for a couple of studies at this conference (Dimitrov) found that a vaginal microbicide with 50% efficacy and 35% usage which produced a modest level of drug resistance (1.1%) would prevent 11% of infections over a ten-year period.

However there were numerous factors that would affect those figures. One finding was that, proportionally, microbicides might make much more difference in low-prevalence settings than in high-prevalence, established epidemics.

The Beckerman paper compared the hypothesised microbicide in female sex workers in two settings, in South Africa and Ivory Coast, and found that while in South Africa more infections would be prevented (1566 versus 571), this would represent a 48% decline in incidence in Ivory Coast but only a 10% decline in South Africa. It was therefore important to predict local impact.

Microbicide efficacy would be severely affected if it produced widespread condom substitution, to the extent of producing increases in infections in some scenarios.

This was because condoms have higher efficacy than microbicides are likely to have, so that a small decline in condom use is not always compensated for by a larger uptake in microbicide use. Condom substitution would have less impact if microbicides were bidirectional and protected the insertive partner too. Clearly if they are unidirectional, and if microbicides lead to heterosexual men deciding to stop using condoms, they are only increasing their own risk.

Boily presented several different scenarios that modelled condom substitution, all assuming the introduction of a vaginal microbicide of 50% efficacy, used 50% of the time. If initial condom use was 30% then, if a microbicide was bidirectional, condom use could drop by 25% before the infection rate started to rise. But if it was unidirectional, then it would only take an 8% drop in condom usage for infections to rise.

Situations of initially high condom use were much more sensitive to condom substitution: if condom use was initially 90%, it would only take a 2% drop in use for infections to start rising, despite the introduction of a microbicide. Audience members commented that this validated the fears of group representing female sex workers, who have managed to enforce this level of condom use in clients. If microbicide use led men to demand more condomless sex, then both worker and client would be more at risk.

Finally, even modest levels of anal sex in a community could make a big difference to microbicide efficacy. In a situation where a vaginal microbicide of 50% efficacy was being used by heterosexual women solely for vaginal sex, then in the absence of anal sex 32% of infections would be prevented over ten years. That would drop to 20% of infections if 5% of sex was anal, and 12% of infections if 10% was. Levels of anal sex this high have been documented in some heterosexual African populations, including in sex workers. Indeed, because anal sex is so much more effective at transmitting HIV, in some circumstances the introduction of a microbicide solely used for anal sex might produce greater declines in HIV incidence.

Ideally, any microbicide should be designed from the start for both vaginal and rectal use, Boily said. It would also add value if microbicides were bidirectional and effective against some STIs. It was important to maximise adherence and minimise condom substitution, and this would require a considerable information and training initiative amongst men who might otherwise pressure women to drop condoms.

Salim Abdool Karim, principal investigator of the CAPRISA tenofovir-microbicide gel study whose results are expected this summer, said that there was data from his trial showing that rectally-used product might protect vaginally and vice versa: significant levels of tenofovir were observed in the rectum if it had been applied vaginally and vice versa.