Modelling the possible public health benefits

A number of studies have modelled the possible HIV-prevention benefits of mass circumcision.

A 2006 modelling study¹ suggested that its widespread adoption throughout Africa could avert up to 5.7 million HIV infections by 2026. The study found that, assuming circumcision reduced the HIV risk to men by 60%, and that risk behaviour did not significantly increase among circumcised men, African countries could begin to see marked reductions in HIV incidence within a few years of promoting adult male circumcision.

This study was published before the announcement of the results from Rakai and Kisumu. The Rakai investigators used the findings from their trial to make further predictions as to the effectiveness of mass circumcision in April 2007.²

They worked out how many more circumcised men there would have to be in order for a number called R₀ to become less than one. R₀ is the so-called 'reproductive number' of an epidemic, meaning the average number of people someone infects with a disease before they die of it. It has to be above one in order for an epidemic to grow.

The Rakai researchers found that the R₀ in areas of Africa likely to benefit from circumcision was currently 1.44. Circumcision programmes would drive it below 1.0, if circumcision had the 50% efficacy seen in the ITT analysis at Rakai, as long as more than three  quarters of currently uncircumcised men were circumcised. Only half the uncircumcised male population would have to undergo the operation if circumcision was 60% efficacious.

According to mathematical modelling done by Kyeen Mesesan of the Yale University department of Public Health,³ circumcision could reduce HIV infections even if there is behavioural disinhibition. Even if a 50% rate of consistent condom use went down to zero there would be a net positive effect on HIV incidence in the population if circumcision was more than 63% efficacious. This, at least, would be the situation as modelled on the population of Soweto, South Africa.

The 35% rate of circumcision seen in Soweto was already protecting some of the population. HIV prevalence over the next 20 years in the area was forecast to increase by 1%, from 16 to 17%. If circumcision did not exist, the model showed, prevalence 20 years from now would be 23%.

If the rate of circumcision was raised from 35 to 55%, then prevalence would decline to 14% in 20 years.

If the protective effect seen in the intent-to-treat analysis of the Orange Farm was the true effect, then condom use could fall to 3% before the benefit of such a 20% increase in circumcision was completely lost.

A mathematical model by Richard White of the London School of Hygiene and Tropical Medicine, presented at the Mexico International AIDS Conference in 2008,⁴ found that tripling the rate of male circumcision in a country with a current circumcision rate of 25% and a high rate of heterosexually acquired HIV would eventually halve HIV incidence.

Nevertheless, it would take 50 years for that 75% rate of circumcision to reach full effectiveness, and it would require circumcision of the majority of sexually active adult men (aged 15 to 45). It was shown that circumcision would immediately result in money being saved, as the cost per HIV infection averted would always be lower than the cost of providing care if that infection had not been averted.

White’s model took a baseline population where HIV prevalence was 18% in men and 25% in women, where HIV prevalence peaked at around 28 in women and 34 in men, and where 25% of the men were already circumcised, and then examined the effect over the next half-century of a national programme that tripled the male circumcision rate to 75%.

He found that there would be an immediate reduction of 8.1% in HIV incidence the second year after circumcision. After ten years, incidence would have gone down by nearly a quarter, but it would take 50 years for circumcision alone (ignoring the effect of any other prevention measures or of HIV treatment) to halve male incidence.

Incidence in women would decline more slowly than in men, going down by only about 12% after ten years, but would catch up as the effect of circumcision spread into the general population, and incidence would have declined by 44% after 50 years.

If circumcisions were performed on boys under 15, before the age of first sex, these effects on HIV incidence would be delayed by 20 years and if performed on babies only, the effects would be delayed by 40 years, as it is much more effective initially to circumcise men at the time they are most vulnerable to HIV.

Including HIV-positive men in a national circumcision programme would make very little difference to incidence reductions, White said, as would the subsequent rate of STIs in the circumcised population (with about 5% of the decline in incidence being due to circumcision reducing herpes infections, rather than a direct effect on HIV incidence).

Yet large-scale behavioural change would have an effect. In White’s model, if 40% condom use in casual and commercial sex declined to 20% (condom use in steady relationships was assumed to be zero), it would wipe out the benefits of circumcision.

At the Mexico conference Nicolai Lohse of the WHO/UNAIDS Male Circumcision Working Group⁵ presented a synthesis of recent models of the effect of circumcision. The group reviewed six previous modelling studies. Most of the models took their input assumptions from models of the high-prevalence heterosexually spread epidemics seen in southern Africa.

A high level of agreement over the main findings was found among the models. Most models agreed that circumcision would produce a reduction of around 60% in the risk of HIV acquisition for men who had the operation.

The risk to women of acquiring HIV would also be reduced if men were circumcised, as there would be fewer HIV-positive men in the population. The risk to women of HIV acquisition would decline by 2% if only 5% of men were circumcised, by 20% if 50% of men were circumcised, and by 38% if 95% were circumcised.

Lohse said that most models were more optimistic than White’s about the effect of ‘condom migration’ on HIV incidence. His synthesis of the models indicated that the risk to women of acquiring HIV would only rise above baseline levels if condom use in men fell by more than two-thirds, while men would benefit from circumcision even if they stopped using condoms altogether.

Lohse’s group published more data from their review in PLoS Medicine in 2009.⁶ They concluded that in the high HIV-prevalence countries of southern Africa, between 5 and 15 men would need to be circumcised to prevent one HIV infection in the ten following years, at a cost of between $150 and $900 per infection prevented.

In countries with a somewhat lower HIV incidence and prevalence, circumcision would have less impact. The group agreed that in such countries, circumcision programmes which focused on specific subpopulations, such as men with HIV-positive partners, men with STIs, soldiers, truck drivers and migrant workers, could have a substantial impact.

The group concluded that women would benefit indirectly from circumcision because their likelihood of meeting an HIV-positive male partner would decline. Moreover, reductions in sexually transmitted infections in both men and women would reduce women’s risk of acquiring HIV.

Two of the models looked at the impact of circumcising men who already had HIV. They concluded that the problem of such men resuming sex too early after their operation would be unlikely to have an impact on a population level because the post-healing time is relatively short. They said that systematic exclusion of men with HIV from circumcision might lead to stigma for all uncircumcised men.

With regard to behaviour change post-circumcision, three models suggested risk compensation by circumcised men and their partners would only have a “small effect” at the population level, unless it was to the extent of complete abandonment of condoms.

The models showed that circumcising men who have not started sexual activity leads to the greatest population-level benefit in the long term, but circumcising 25- to 34-year olds has the biggest benefit in the first 20 years.

A study presented at the 2006 International AIDS Conference⁷ found that assuming a cost per circumcision of $55.75, the cost of a mass male circumcision programme where HIV prevalence was 8.4% - regardless of the programme’s actual coverage - would be $550 per HIV infection averted. When antiretroviral and other treatment costs were taken into account the efficacy of circumcision would be a net saving of $753 per circumcision; circumcision would be a cost-saving measure at any efficacy above 21%.

In contrast, a similar analysis done by the Rakai researchers in Uganda⁵ found that the costs per infection averted would be $1269 to $3911, depending on the efficacy of circumcision for either or both sexes, assuming 75% service coverage. A model by Bertran Auvert, lead researcher in the Orange Farm study, and others was presented to the Sydney International AIDS Society conference in 2007.⁸ It found that although rolling out circumcision in the 14 countries would cost the public sector between $315 million and $532 million during the first five years, a much greater cost would be borne by the private sector, particularly by individuals paying fees for circumcision operations. Private expenditure on circumcision might top $1.2 billion. Although considerable investment would be required in the first five years of a circumcision roll-out programme, Auvert calculated that over 20 years between $3 billion and $4 billion would be saved in HIV treatment and care costs.

He also calculated that the cost of circumcision per infection prevented would be between $113 and $375 and that, in the first 20 years, it would be necessary to circumcise between four and nine men to prevent each new HIV infection.

In the 2008 model presented by Richard White,⁴ the cost of averting one HIV infection, for the base scenario in which 15 to 45 year olds were circumcised, would be $1806 in the first two years, but would decline to only $200 by the time the programme had been running for 20 years. This compares with a cost of over $4000 for treating the HIV infection that would otherwise have happened, so circumcision would be cost saving from the start if performed on adult men.

Another mathematical model presented at Sydney⁹ showed that universal circumcision would have the greatest impact on HIV incidence, but that targeting circumcision at men with the most sexual partners and those aged 20 to 30 years old would be the most effective way of reducing HIV prevalence.

Gregory Londish of the University of New South Wales, Sydney found that complete male circumcision in an average sub-Saharan country could reduce HIV prevalence in 2020 from 8.3 to 5.3% and annual incidence from 1.35 to 0.73%.

Targeting only 20- to 30-year-old men or men with greater sexual activity produced the most cost-effective reduction in HIV prevalence, 2.0% and 1.1% respectively. But the benefits would be smaller with increasing sexual activity in men who have been circumcised, and would be completely eliminated if more than 40% of circumcised men increased their sexual activity.

The UNAIDS/WHO review of models published in 2009⁶ found that each infection that was prevented because of circumcision would cost between $150 and $900, calculated over a ten-year time period. When calculated over 20 years, the cost per prevented infection was $100 to $400. Costs would be higher in lower prevalence countries.

These costs are based on $30-$60 per adult circumcision, and a life-time treatment cost of $7000 per HIV infection (first-line therapy only).

The UNAIDS/WHO review did not find that circumcising newborn babies would be cost-effective. Although circumcision at this stage is safer and cheaper, the impact on HIV would not be seen for over 20 years.

In contrast, a more recent analysis of the cost-effectiveness of circumcision in Rwanda, where adult HIV prevalence is 3%,¹⁰ has concluded that circumcising newborn babies would be cheaper and prevent more infections than providing the operation to adolescents or adults.

The researchers developed a cost-effectiveness model and applied it to three hypothetical groups of Rwandans: newborn boys, 15-year-old boys, and 30-year-old men. In contrast to other studies, cost savings were considered over the lifetime of the circumcised males, rather than the next 10 or 20 years.

The cost-per-circumcision was estimated to be much lower for babies (US$15) than for adolescents or adults (US$59), because the adult operation is more complex, requires anaesthetic and other more expensive materials, and involves HIV testing and counselling, as well as more staff time and greater infrastructure.

The researchers estimated that if 150,000 men aged 30 were circumcised in 2008, then 859 infections would be avoided in their collective lifetimes. The impact would be seen in the years immediately following the operation.

Circumcising the same number of 15-year-old boys would prevent 1283 HIV infections, with the greatest reduction in infections taking place between 2018 and 2037 (i.e., when the boys are aged 25 to 44).

If babies were circumcised within a month of birth, a similar number of infections would be avoided: 1288, with the greatest benefit seen between 2033 and 2052.

Providing the operation to babies would cost $2.25 million and would save $3.80 million in future costs, and would therefore be cost saving. However, as the operation for adolescents or adults is considerably more expensive ($8.85 million for each group), the costs are higher than the savings.

The researchers recommend that countries with moderate HIV epidemics should offer routine circumcision to newborn babies, integrated into existing health services. In addition, adolescents and higher-risk adults should be offered circumcision until the ageing of boys circumcised at birth makes such a policy obsolete.

They called on policymakers to stop treating HIV solely as an emergency issue. Although the benefit of infant circumcision will not be seen for another two decades, they say, “action cannot be deferred simply because gains will be in the distant future.”

Despite the convincing scientific evidence of its efficacy, one model, comparing the impact of circumcision with that of scaling up HIV treatment and/or improving condom usage, casts doubt on whether, in situations of limited resources, mass circumcision programmes are the most effective and cost-effective way of reducing HIV prevalence and incidence.

In preliminary results from a mathematical model set up by researchers from the British Columbia Centre for Excellence in HIV/AIDS,¹¹ male circumcision was found to have a considerably lower impact than condom use or antiretroviral therapy (ART) coverage on new HIV-infection rates and on death rates in men in South Africa. The model compared simulated scenarios involving various levels of male circumcision, condom use and ART coverage up to 2025.

The simulations inputted male circumcision rates at 51 (the current level), 75 and 90%; condom-use rates at 14% (current level), 50, 75, 80 and 90%; and ART coverage, starting at a CD4 count of 200 cells/mm³, at 21% (current level), 50, 75, 80 and 90%.

For the purposes of this initial, simple version of the model, both the number of women infected with HIV and their infectiousness remained the same. In a more sophisticated model which includes both male and female infections, reductions in female infections and infectiousness would have a synergistic effect, which is currently being calculated.

The model found that scaling condom use up from 14% (the current estimated level) to 50% and antiretroviral therapy coverage from 21 to 80% would result in an estimated 950,000 infections averted in men aged 15 to 50 by 2019. Raising ART use to 50% and condom use to 80% would have a similar effect.

Raising both to 50% would result in 700,000 fewer infections. But raising circumcision rates from the current 51% of men circumcised to 90% would only add another 48,000 infections averted to this total.

Condom use and ART coverage, alone or in combination, were found to reduce new HIV infections by between 64 and 95% by 2025 and to reduce mortality by 10% to 34%. Circumcision brought about a 3 to 13% reduction in new HIV infections and a 2 to 4% reduction in mortality.

According to researcher Vivian Lima, its impact was overshadowed when combined with the other interventions. “We were surprised by how little effect it had,” she said.