The authors have declared that no competing interests exist.
Conceived and designed the experiments: MH. Performed the experiments: MH. Analyzed the data: MH. Wrote the first draft of the manuscript: MH. Contributed to the writing of the manuscript: MH MG NF. Agree with the manuscript’s results and conclusions: MH MG NF. All authors have read, and confirm that they meet, ICMJE criteria for authorship.
Empirical studies and population-level policy simulations show the importance of voluntary medical male circumcision (VMMC) in generalized epidemics. This paper complements available scenario-based studies (projecting costs and outcomes over some policy period, typically spanning decades) by adopting an incremental approach—analyzing the expected consequences of circumcising one male individual with specific characteristics in a specific year. This approach yields more precise estimates of VMMC’s cost-effectiveness and identifies the outcomes of current investments in VMMC (e.g., within a fiscal budget period) rather than of investments spread over the entire policy period.
The model has three components. We adapted the ASSA2008 model, a demographic and epidemiological model of the HIV epidemic in South Africa, to analyze the impact of one VMMC on HIV incidence over time and across the population. A costing module tracked the costs of VMMC and the resulting financial savings owing to reduced HIV incidence over time. Then, we used several financial indicators to assess the cost-effectiveness of and financial return on investments in VMMC. One circumcision of a young man up to age 20 prevents on average over 0.2 HIV infections, but this effect declines steeply with age, e.g., to 0.08 by age 30. Net financial savings from one VMMC at age 20 are estimated at US$617 at a discount rate of 5% and are lower for circumcisions both at younger ages (because the savings occur later and are discounted more) and at older ages (because male circumcision becomes less effective). Investments in male circumcision carry a financial rate of return of up to 14.5% (for circumcisions at age 20). The cost of a male circumcision is refinanced fastest, after 13 y, for circumcisions at ages 20 to 25. Principal limitations of the analysis arise from the long time (decades) over which the effects of VMMC unfold—the results are therefore sensitive to the discount rate applied, and more generally to the future course of the epidemic and of HIV/AIDS-related policies pursued by the government.
VMMC in South Africa is highly effective in reducing both HIV incidence and the financial costs of the HIV response. The return on investment is highest if males are circumcised between ages 20 and 25, but this return on investment declines steeply with age.
Medical male circumcision reduces incidences of HIV. Gorgens and colleagues present a new model that identifies the best age for this treatment, in terms of cost efficiencies and disease incidence.
Every year, about 2 million people become infected with HIV, the virus that causes AIDS. Although antiretroviral therapy can keep HIV in check, there is no cure for AIDS. Consequently, prevention of HIV transmission is an important component of efforts to control the AIDS epidemic. Because HIV is most often spread through unprotected sex with an infected partner, individuals can reduce their risk of becoming infected by abstaining from sex, by having only one or a few sexual partners, and by using male or female condoms. In addition, three trials undertaken in sub-Saharan Africa a decade ago showed that male circumcision—the surgical removal of the foreskin, a loose fold of skin that covers the head of the penis—can halve the sexual transmission of HIV from HIV-positive women to HIV-negative men. In 2007, the World Health Organization (WHO) and the Joint United Nations Programme on HIV/AIDS (UNAIDS) recommended that individuals living in countries with generalized HIV epidemics (countries where more than 1% of the general population is HIV positive) should be offered voluntary medical male circumcision (VMMC) to help prevent HIV transmission.
In 2011, following several studies that assessed the population-level effectiveness and cost-effectiveness of VMMC scale-up, WHO/UNAIDS set a target of having at least 80% of all men aged 15–49 circumcised in 14 priority countries in Africa by 2015. Good progress towards this target was made, but VMMC scale-up needs further improvement. Scenario-based studies provide estimates of the overall costs and impacts of VMMC policies over the whole policy period, which can span several decades, but policy makers sometimes need to know the returns on investments in VMMC over the current financial year. Here, the researchers use an “incremental” cost-effectiveness approach (one that analyzes the expected consequences of circumcising one male with specific characteristics in a specific year) to provide estimates of the impacts of current VMMC investments and to obtain more precise estimates of VMMC’s cost-effectiveness for HIV prevention in South Africa, a country with a generalized HIV epidemic.
The researchers developed a mathematical model consisting of three elements: a demographic and epidemiological model of the South African HIV epidemic adapted to analyze the impact of one VMMC on HIV incidence over time and across the population; a module to track the costs of VMMC and financial savings from HIV infections averted by VMMC; and several financial indicators to assess the cost-effectiveness of and financial return on investments in VMMC. Using this model, the researchers estimate that the circumcision of a young man up to age 20 prevents, on average, 0.2 HIV infections, and that the number of infections averted by VMMC declines steeply with age at circumcision. Estimated financial savings from one VMMC at age 20 are US$617 at a discount rate of 5% (discounting translates future costs and benefits into present-day values); VMMC at younger and older ages results in lower financial savings. From a purely financial perspective, circumcisions at age 20 are most effective with a financial return of 14.5% (they are equivalent to a financial investment yielding an interest of 14.5% annually). Finally, the costs of a circumcision are refinanced (covered by reductions in the costs associated with HIV infection) fastest for circumcisions at ages 20–25.
Because the effects of VMMC unfold over many decades, the accuracy of these findings is likely to be affected by the future course of the AIDS epidemic, by HIV/AIDS policies adopted by the South African government, and by the discount rate applied in the cost-effectiveness analysis. Moreover, some of the assumptions made by the researchers in their model may affect the accuracy of their findings. Overall, however, these findings, which provide estimates of the consequences of current VMMC policies, suggest that VMMC in South Africa (and possibly in settings where a lower proportion of the population is HIV positive) is a highly effective HIV prevention intervention that is cost-saving under many circumstances. These findings also indicate how age at circumcision affects the impacts of circumcision. Specifically, they suggest that the return on investment is highest if males are circumcised between ages 20 and 25 but declines steeply with increasing age.
This list of resources contains links that can be accessed when viewing the PDF on a device or via the online version of the article at
Information is available from the US National Institute of Allergy and infectious diseases on
Information is available from
The World Health Organization provides information on all aspects of
The effectiveness of male circumcision in reducing sexual transmission of HIV from HIV-positive women to HIV-negative men has been generally accepted at least since 2007 [
These programmatic developments have motivated, and were informed by, a number of studies assessing the population-level effectiveness [
Four general lessons emerge from this academic literature. First, VMMC—in countries with a high burden of HIV and low prevalence of male circumcision—is very effective in averting HIV infections among men. Njeuhmeli et al. [
The drive toward expanding coverage of male circumcision has resulted in a steep increase in the number of male circumcisions performed. In 2014, 3.2 million men in the 14 priority countries were circumcised, contributing to a total of 9.1 million circumcisions since 2008 [
The literature on the experience of the scale-up of VMMC so far offers lessons for improving the efficiency of VMMC services, by analyzing cost drivers [
This paper complements other current efforts to estimate the impacts of male circumcision by age at circumcision but differs in terms of the approach. Rather than assessing the overall costs and impacts of a VMMC policy scenario (specifying the number of circumcisions at different ages performed over some policy period), it estimates the impacts over time and across the population of circumcising one male individual at a specific age in a specific year. The present analysis is related to scenario-based approaches as both rely on a fully specified epidemiological model to generate projections of the state of HIV/AIDS and the effects of the policy contemplated.
However, the purposes and outcomes of the respective analyses differ. Policy scenarios are an effective and intuitive way of describing the costs and consequences of policies under consideration and obtaining some measures of cost-effectiveness (e.g., costs per HIV infection averted over the policy period).
The value added from the incremental analysis arises in two areas, but both are related as they address challenges associated with the long time frames required to capture the effects of investments in VMMC—typically, policy scenarios therefore extend over several decades. In contrast, the incremental approach provides estimates of the impacts of current VMMC investments—e.g., informing policy makers on what the returns are on spending considered as part of the next annual government budget.
Additionally, the incremental analysis offers more precise estimates of the cost-effectiveness of current VMMC policies, for two reasons. First, policy scenarios show average effects of interventions occurring over (possibly long) policy periods and, therefore, do not offer precise estimates of the effects of interventions at any point in time. Second, policy scenarios are subject to a cut-off problem, as some of the effects of interventions studied within the policy period lie outside the policy period. In particular, they barely capture the effects of a VMMC performed late in the period. The incremental analysis gets around these problems by focusing on VMMC policies in a particular period (so there is no averaging over periods), and because it can follow the impacts of one VMMC until it dissipates (so there is no arbitrary cut-off) and capture the full effects.
The model used in this analysis consists of three elements: a demographic and epidemiological model used to estimate the projected impacts of one VMMC on HIV incidence, a costing module used to project the costs caused by a new HIV infection (and the savings from HIV infections averted as a consequence of one VMMC), and a financial analysis describing the cost-effectiveness of and the financial returns to investments in VMMC.
The analysis of the effects of VMMC on HIV incidence was conducted using an adapted version of the ASSA2008 national (“lite”) model [
We chose this model, because it allows an incremental analysis such as the one performed here in a relatively straightforward fashion. The ASSA2008 model, however, does not explicitly account for male circumcision and therefore needed to be adapted in four steps (see
To translate the estimates of the impacts of VMMC on HIV incidence into projections of the resulting changes in the demand for HIV/AIDS services, an epidemiological model describing disease progression and transition into treatment and a costing framework were required. The ASSA2008 model has not been used for this purpose, because it is known to underestimate survival on antiretroviral treatment [
The costs of treatment, which are the dominant aspect of the costs caused by HIV infections, were updated in line with recent estimates produced for the South African HIV/TB investment case [
Capturing the effectiveness and cost-effectiveness of VMMC is complicated, because the effects of VMMC on HIV incidence are spread over a long period (decades), and the cost savings from reduced HIV incidence are spread out even further (another several decades). Estimates of the most common indicators (net costs of or savings from one VMMC) are therefore highly sensitive to the discount rate applied, and the main results (applying a discount rate of 5%) are complemented by a sensitivity analysis with regard to the interest rate. Additionally, the analysis offered some indicators more commonly used for the analysis of financial investments—the amortization period (for a given interest rate, how long does it take until the initial costs are refinanced?) and the financial rate of return (up to which interest rate are the financial returns positive, and is the investment in VMMC a good financial investment?).
To estimate the consequences of VMMC occurring in 2013, the number of male circumcisions performed at a specific age in 2013 was raised by one, at age 0 (for early infant male circumcision) and in 5-y increments from age 10 to age 55. The analysis focuses on the consequences of circumcising South African HIV-negative individuals irrespective of ethnicity or geographic location; i.e., it was assumed that the individual circumcised, at each age, was randomly drawn from HIV-negative individuals from the four risk groups at that age. As the HIV status of an individual undergoing circumcision may be unobserved, the analysis was repeated assuming that individuals were drawn randomly from all male individuals at a specific age, regardless of HIV status. The model was then used to project the expected consequences for the individual circumcised and across the population.
The adapted model includes some population-level assumptions on the uptake of male circumcision (see
We determined the impact of VMMC on HIV incidence for the individual circumcised and among men and women across the population, identified financial savings accruing from VMMC by age, and incorporated these savings in the analysis of the cost-effectiveness of VMMC, by age of the individual circumcised. The savings are spread over a very long period, with calculations of potential savings sensitive to the discount rate applied. Because of the large magnitude of the financial savings, we also interpreted investments in VMMC as financial investments, applying indicators like the financial rate of return and amortization period.
Male circumcision reduces HIV incidence both directly and indirectly—directly, because it reduces the HIV infection risk for the uninfected individual circumcised, and indirectly, because the reduced infection risk for the individual circumcised also reduces the risk that he transmits HIV to sexual partners and so on. The direct effects are summarized in
For an uncircumcised 15-y-old male in 2013, the probability that he will contract HIV in a specific future year steeply increases to 2.1% by age 25, reflecting increasing sexual activity, a shift to higher-risk sexual behavior, and an increasing risk of contact with HIV-positive females. Subsequently, it declines steadily to 1.3% by age 30, 0.4% by age 40, and 0.1% by age 50, reflecting declining sexual activity (for men from age 27) and the changing composition of men not infected with HIV (men reaching older ages without becoming infected are statistically more likely to have adopted, and to adopt, low-risk sexual behavior). For a circumcised male, the probability that he will contract HIV peaks at a much lower rate (1.2%) at age 26. For infant circumcision (male newborns circumcised within a year of birth), the lifetime HIV infection risk is similar to that for a 15-y-old, but occurs with a long delay once individuals become sexually active.
Overall, each male circumcision averts about 0.09 HIV infections over the course of a circumcised individual’s life for circumcisions at ages 15 or 20 (
To determine the effectiveness of VMMC in reducing new infections, and to guide the design of VMMC programs, it is important to know how many men who might undergo VMMC are HIV-positive already, and how this might compromise the effectiveness of VMMC. If HIV prevalence among males undergoing VMMC is the same as for their cohort overall, the effects of VMMC are not much diminished at young ages (up to age 20), but decline by 10% by age 25 and by over a quarter for circumcisions at ages 35 or 40.
The HIV incidence-reduction effects of VMMC, though, are larger than the direct effects (on individuals circumcised), because individuals not infected also do not transmit HIV. The indirect effects could potentially be large because of relatively high rates of male-to-female transmission of HIV (in 2014, 55% of estimated adult HIV infections in South Africa occurred among women [
The magnitude of such indirect effects is illustrated in
Note: "Male (direct effect)" shows the expected effect on the individual circumcised, "Females" shows the population-level effects among female partners, "Other males" represent population-level effects among males other than the individual circumcised, and "MTCT" stands for HIV infections averted through mother-to-child transmission.
Overall, each male circumcision for young adults (age 20) averts 0.23 HIV infections, of which 0.09 can be attributed to the direct effects and 0.14 HIV infections to the indirect effects (
Note: Lines are stacked, so that the distance between the lines gives the effect for each group, and the highest line shows the total effect.
The costs caused by one new HIV infection are illustrated in
Results on the cost-effectiveness of male circumcision follow from the estimates on the effectiveness and the unit costs of VMMC. If one adult VMMC costs US$104, then circumcisions between ages 10 and 20 avert HIV infections at a cost of US$450 to US$478 per infection averted (
Age at male circumcision | Impact on HIV Incidence | Male circumcisions per HIV infection averted | Cost per HIV infection averted (US$) | Net savings from one male circumcision (US$, at 5% disc.) | Amortization period (y, at 5% disc.) | Financial rate of return (%) | |||
---|---|---|---|---|---|---|---|---|---|
Total | Direct | Indirect | (at 0% disc.) | (at 5% disc.) | |||||
0 | 0.236 | 0.082 | 0.154 | 4.2 | 221 | 859 | 301 | 30.6 | 7.8 |
10 | 0.232 | 0.084 | 0.147 | 4.3 | 450 | 1,157 | 401 | 22.8 | 9.9 |
15 | 0.218 | 0.085 | 0.133 | 4.6 | 478 | 867 | 475 | 17.4 | 11.5 |
20 | 0.227 | 0.088 | 0.139 | 4.4 | 460 | 659 | 617 | 12.2 | 14.5 |
25 | 0.159 | 0.083 | 0.076 | 6.3 | 657 | 854 | 412 | 11.6 | 13.6 |
30 | 0.079 | 0.059 | 0.020 | 12.6 | 1,316 | 1,749 | 123 | 16.6 | 8.8 |
35 | 0.040 | 0.034 | 0.006 | 25.1 | 2,623 | 3,455 | −2 | n.a. | 4.9 |
40 | 0.022 | 0.020 | 0.002 | 46.2 | 4,826 | 6,135 | −52 | n.a. | 1.8 |
45 | 0.014 | 0.013 | 0.000 | 72.7 | 7,592 | 9,268 | −74 | n.a. | −0.9 |
50 | 0.009 | 0.009 | 0.000 | 113.3 | 11,819 | 13,709 | −86 | n.a. | −3.6 |
55 | 0.005 | 0.005 | 0.000 | 214.2 | 22,350 | 24,157 | −95 | n.a. | −6.8 |
These estimates point to the potential of male circumcision to contain the costs of the HIV/AIDS response and contribute to financial sustainability of the HIV response: the one-off costs per HIV infection averted through VMMC at younger ages are of a similar magnitude as the annual cost of treatment (between US$267 and US$531 as of 2014). Considering that people living with HIV in South Africa (provided they initiate treatment early) can have a near-normal life expectancy [
However, because the impact of VMMC on HIV incidence is distributed over a long period (about 20 y for young adults, but somewhat less at older ages; see
For male circumcisions occurring above age 20, the relative magnitude of the projected savings from one VMMC (
However, for any discount rates commonly applied in policy analysis, the longer lags mean that the financial savings for circumcisions at young ages carry a lower weight. If a discount rate of 5% is applied, the financial savings from infant circumcision are about one-half lower than for circumcisions at age 20, even though the undiscounted cost savings are 20% higher.
Indeed, owing to the extremely long period over which the effects are spread, the discount rate applied becomes one of the most important factors in interpreting the financial savings (
This sensitivity analysis with respect to the discount rate also offers insights on the robustness of findings with respect to treatment costs (the dominant factor in financial savings). If unit costs of treatment gradually decline by one-half by 2050, i.e., by about 2% annually, the implications are very similar to an increase in the discount rate of two percentage points—net savings are much diminished, but results on the cost-effectiveness of VMMC overall and the role of age at circumcision come out similarly.
Nevertheless, the thrust of the findings—that VMMC generates very substantial financial savings which more than offset the costs of VMMC—is robust to changes in the discount rate. For VMMC at ages 20 or 25, the costs of VMMC are amortized (i.e., fully refinanced by the resulting savings) after only 12 y; over this period, the discount rate does not play a dominant role (
VMMC in South Africa is an intervention that is highly effective in preventing HIV infections, and also a very good financial investment, with savings far outweighing the costs for circumcisions at most ages. Each male circumcision averts up to 0.23 HIV infections for young adults (up to age 20), but the effect declines for VMMCs performed at higher ages, as the reduction in the lifetime risk of contracting HIV for the individual circumcised diminishes, and even more so because the indirect effects decline steeply with increasing age at circumcision. For circumcisions at younger ages, the effect is similar to circumcisions at age 20, but the effects occur with a long delay, especially for EIMC.
In terms of financial returns, circumcisions at age 20 are most effective, with a financial rate of return of 14.5% (
In contrast to other studies, which analyze and compare various population-level scenarios, we conducted an incremental analysis to determine the expected effects of circumcising one male individual at a specific age. Our analysis thus did not provide estimates on how a VMMC policy contributes to attaining national goals on reducing HIV incidence on the population-level over some policy period. Instead, our approach yields more precise estimates of the effects and cost-effectiveness of VMMC and provides direct estimates of the consequences of current policies under consideration.
The improved precision is related to the long time frames (VMMC to HIV infections averted, and costs caused by new HIV infections which also extend over decades). National policy scenarios address this problem by adopting long time horizons (now frequently up to 2050), but produce imprecise estimates of cost-effectiveness as inputs and outcomes are aggregated over long periods and are subject to bias (especially for circumcisions occurring late in the policy period, the consequences are barely captured). The incremental analysis offered here works around these problems by considering the effects of a policy at a specific point of time only, which then allows us to track the consequences of this policy as long as necessary to capture the full effects (
In terms of the quantitative results, the findings from our analysis on the cost-effectiveness of VMMC are broadly consistent with those of other studies on VMMC in general or across various age groups: VMMC is a highly effective HIV prevention intervention in a country with a generalized HIV epidemic and under many circumstances is cost-saving as the financial savings owing to reduced HIV incidence outweigh the costs of providing VMMC. VMMC is most effective at about age 20, declines at higher ages because of the diminished direct and indirect effects on HIV incidence. For VMMC at lower ages, the lifetime impact on HIV infections averted is about the same as for young adults. However, because of the delayed impact (especially for EIMC), the effects of circumcising infants or young adolescents are diminished relative to young adults when some discounting is applied to the outcomes.
With regard to informing policies, the relevance of our approach derives from the ability to make precise statements of the effects of policies considered under the government’s current budget process, i.e., typically over a one-year period. In this regard, it does not only offer estimates of the health returns to investments in VMMC but also of the implications of current policies in terms of the fiscal space absorbed by the HIV/AIDS response—what are the net costs or net savings over the coming years associated with current spending on VMMC, and—if there are net savings—how quickly are the current outlays refinanced by anticipated savings?
More fundamentally, the approach also offers lessons relevant for the design of the HIV/AIDS program. What is the contribution of current or additional investments in VMMC to the long-term costs and the financial sustainability of the HIV/AIDS response? Can outcomes or long-term costs be improved by reallocating current funds between interventions [
To some extent, the analysis and the magnitude of the estimated effects of VMMC is specific to South Africa. The magnitude of the financial savings, however, suggests that VMMC would save costs at a much lower level of HIV prevalence than observed in South Africa, at least for some age groups. Our analysis identified some aspects of the cost-effectiveness of VMMC, which are plausibly relevant more generally, notably the decline in the effectiveness of VMMC by age, reflecting three factors: (1) at higher ages, controlling for behavior, the remaining lifetime risk of contracting HIV is lower; (2) a sorting effect: individuals adopting high-risk behavior are more likely to become infected early, whereas individuals not infected at higher ages are statistically more likely to have adopted and to adopt low-risk sexual behavior; and (3) the role of declining sexual activity and time lags between successive rounds of downstream infections, which explain much of the decline in the indirect effects of VMMC.
The principal limitations of the analysis arise from the long time (decades) over which the effects of VMMC unfold. This is evident from the sensitivity analysis with respect to the discount rate (
In terms of the modelling assumptions, perhaps the most significant factor introducing uncertainty to the projections is the assumed zero impact of male circumcision on male-to-female transmission of HIV. If an effect of VMMC on male-to-female transmission of HIV is present, the indirect impact of male circumcision on female sexual partners would come out larger (and could even double by some account [
(DOCX)
(PDF)
(PDF)
(DOCX)
The paper benefited from a discussion of an earlier draft with experts convened by the US Agency for International Development (USAID), and comments by Emmanuel Njeuhmeli, Jason Reed, Steven Forsythe, Delivette Castor, and Anthony Kinghorn. In South Africa, the work benefited from inputs and support from Debbie Bradshaw, Susan Cleary, Teresa Guthrie, Eva Kiwango, Lusanda Mahlasela, Gesine Meyer-Rath, and Helen Odido. Michael Obst provided research support. The research reported in this manuscript has been prepared by Markus Haacker, Marelize Gorgens, and Nicole Fraser-Hurt as part of their employment with the World Bank and therefore its publication must comply with the World Bank Open Access Policy (
early infant male circumcision
gross domestic product
President’s Emergency Plan for AIDS Relief
Joint United Nations Programme on HIV/AIDS
US Agency for International Development
voluntary medical male circumcision