This study provides the first estimates of the cost of scaling up MAMC in sub-Saharan Africa, the number of circumcisers needed and the likely savings due to averted HIV medical care costs. After numerous observational studies, three randomised controlled studies, a modelling study and a cost-effectiveness study, this analysis provides further evidence supporting the rapid roll-out of MC in sub-Saharan Africa. Cumulative net costs at 20 years are negative for the base case and for almost all sensitivity analyses, indicating that the intervention will save money.
This study has some limitations. On the costing/demand aspect, the predicted number of circumcisers required depends on the assumed period during which full MAMC scale-up is achieved. Our base case assumption that a period of 5 years will be sufficient to circumcise most of the uncircumcised males might be seen as optimistic. The sensitivity analysis showed that the required number of circumcisers would be less with a longer start-up phase; however, the associated costs to scale up MAMC were relatively insensitive to this duration. Our assumption that a high fraction of uncircumcised men would accept circumcision may be considered an upper limit, reflecting the hypothesis that education campaigns promoting MAMC might even increase demand above levels reported in acceptability studies. Finally, our assumption about circumciser productivity is based on recent experience and may turn out to be conservative as efforts are made to design and evaluate more efficient surgical methods and ways of deploying staff.
Despite the consistent impacts found in the 3 RCTs, the exact impact that MAMC will have when scaled up in the field remains uncertain. For example, impact observations in these trials were all limited to the first 2 years after surgery. In addition, the indirect effect of circumcision on preventing mother-to-child transmission and associated reductions in medical costs were not taken into account in our model.
The mathematical model used to calculate epidemiological impact was a simple susceptible-infected compartmental model 
, which, for example, does not account for heterogeneity in HIV transmission by age. Our prediction of impact on HIV infections is similar to that of other HIV epidemic models 
. For example, our estimates of the number of MAMCs required to avert an HIV infection and program cost per HIV infection averted are consistent with a prior analysis using slightly different modelling assumptions 
. Our impact estimate is higher than one study 
, with differences due mainly to epidemic severity and modelling time horizon. This consistency with past work, in combination with the robustness of our findings in multivariate sensitivity analysis, suggests that our results are likely to be reasonable predictions of epidemic impact and associated financial savings from HIV care averted.
Our analysis cannot be used to estimate precisely the difference in costs between a private and a public scenario, due to a lack of firm data on the cost of either scenario, especially with evidence of wide variation in private pricing. A pragmatic view suggests that each delivery approach has advantages and disadvantages. The main advantage of the private provider scenario is that it is immediately available: subsidies to private sector MAMC facilities by public or private donors will make this sector quickly operational. The disadvantage of the private sector approach is its likely higher long-term cost, the lack of private doctors in many rural areas threatening geographical equity in access and the insufficient number of doctors to fully cover the need for MAMC.
The main advantage of the public sector approach is its potentially lower cost in the longer term and potentially better geographic equity. The main disadvantage is that the health system may take time to make MC available on a large scale. The public system may require infrastructure development and training of health workers.
Summed over all the countries evaluated, the cost estimates for rolling out MAMC may appear to be high. We nevertheless think that this cost is affordable, for several reasons. First, the cost is high only for a few initial years: once most men have been circumcised, the cost will be reduced to the circumcision of men becoming adults (and eventually to newborns). Second, the cost is an investment which will prevent spending far greater resources in treating persons with HIV/AIDS in future years. Compared with our ART costing assumptions, the costs of ART may even increase if treatment initiation criteria are widened (to earlier stages of infection/disease) in coming years, due to longer therapy and an increasing need for more expensive second- and third-line regimens. Furthermore, spending for MC is modest compared with overall HIV control efforts: our prediction of annual funding required for MAMC roll-out for a 10-year period in the public provider scenario is only one-quarter of the current spending of the PEPFAR program: $433 million annually in 5 countries of Southern Africa 
. The projected funding requirements for MAMC represent a significant and highly variable proportion of a country's total public and private health expenditures, estimated at 0.3%, 1.1%, 2.3% and 6.0% for South Africa, Tanzania, Zimbabwe and Botswana, respectively 
. While this has important implications for planning and budgeting, it does not reflect on long-term affordability, since MAMC is cost saving.
We calculated that, over the first 5 years, about 1/4 full-time circumcisers would be required per 10
000 adults. Current general practitioners may be too few (especially those trained and willing to perform MC) and too busy to fulfil such a need 
. Furthermore, the training of general practitioners takes time and it is not reasonable to assume general practitioners will do just MC. Thus we believe that the training of nurses with an accreditation system could be a rapid way to increase the capacity of the private sector. The workforce shortage being the biggest barrier to roll out of MAMC 
, the involvement of nurses is likely to be a crucial step for an accelerated roll-out of MC. It will require some regulation adjustments because in many countries where MC is not common nurses are not allowed to perform MAMC, even in places where traditional circumcisers without medical knowledge and training are tolerated.
One of the main obstacles to the roll-out of MAMC is the relative technical difficulty of the surgery, which requires precise incisions, haemostasis and sutures. The roll-out of MAMC could be greatly facilitated and accelerated by the development of simplified, bloodless methods 
, which would lighten and shorten the training required for health workers and decrease costs. A full review of these bloodless methods and their acceptability in modern medical practice is therefore an urgent public health need 
. In addition, we are exploring the applicability of “task specialization” team methods with substantially higher productivity per circumciser, similar to those pioneered for cataract removal surgery in Asia in the 1980s 
. This approach can also make excellent use of low level health care workers for the less technical parts of the procedure (e.g., patient preparation and wound dressing).
A major consideration in scaling up MAMC is whether to concentrate on a horizontal or vertical approach. The horizontal approach is exemplified by the integration of MAMC into routine clinical practice. It is best represented in this analysis by the private sector scenario, since many general practitioners are likely to do MAMC as part of their varied clinical activities. The vertical approach makes MAMC a stand-alone activity. The public sector scenario may work with a vertical or horizontal emphasis, or a mix. The vertical approach offers potential to contribute uniquely to a rapid scale-up and the horizontal approach offers more structure for sustainability. We believe that a combination is preferred and that the optimal scale-up methods will depend on the health care system settings. We have not distinguished the cost of the public sector's horizontal versus vertical approaches, which will be the subject of future analyses.
The rapid implementation of MC will necessitate more than just funding. It will require broad involvement from many groups: national political leaders, activists, teachers, street leaders, churches and health workers. MAMC roll-out will also require strong and steady political support. The political involvement of South Africa will be key, as South Africa represents a high fraction of the population that could benefit the most and has a leading political role in the African region. Our hope is that the research done in the past 20 years regarding the potential of MC to reduce the spread of HIV will be recognized not merely as scientific progress, but as the foundation for an effective transition from knowledge to high-impact practice.