The magnitude and the impact of HAART treatment and screening, on cervical cancer mortality in HIV-positive women in sub-Saharan Africa remain unknown. The ethical and practical complexities of potentially denying care to patients and following women for their lifetime respectively make it unfeasible to conduct a study to quantify this impact. Computer-based simulation models, however, provide an alternative means of quantifying the potential impact of these interventions. These models are of even greater importance in sub-Saharan Africa, where real-life estimation is hampered by limitations in the long-term follow-up of patients, in cancer diagnosis and in the determination of cause of death.
In this paper, we used a mathematical model to project that mortality due to cervical cancer in HIV positive women in Africa is potentially very high. While this mortality can be worsened by providing HAART without screening, screening can be associated with non-negligible reductions in mortality from cervical cancer. These data confirm and quantify the potential gains of cervical cancer screening in HIV-positive women in regions with high cancer incidence and mortality such as Cameroon. With an estimated 200,000 HIV-infected women in Cameroon 
, our projections imply that screening these women once at HAART initiation would prevent close to 763 deaths due to cervical cancer, while screening these women once at age 35 could prevent approximately 990 deaths in these women.
WHO guidelines for screening in resource-limited settings include at least a one-time screening in the third or fourth decade of life 
. Our projected potential gains with one-time screening in HIV-positive women are much smaller than those in other settings where screening is systematic and more frequent. For example, screening in the UK was estimated to prevent one death for every 65 women screened systematically 
. The reductions in cervical cancer mortality associated with a single screen in HIV-positive women were also less compared to the estimated 25–36% reductions in lifetime risk reported in models of HIV-negative women in five other developing countries 
. These differences could be due to a higher incidence and faster progression of precancerous lesions in HIV-positive women compared to HIV-negative women 
Although we show potential benefits of screening, our analysis did not take the cost of screening into account. A formal assessment of the cost-effectiveness of screening in this and other settings will provide additional information regarding resource needs and potential effects for cervical cancer screening. Our data show that even one time screening at HAART initiation or at age 35 would potentially be beneficial. According to guidelines from the Commission on Macroeconomics and Health, a policy is generally considered cost-effective if the incremental cost per life saved is less than the country's GDP per capita 
. Our analyses estimate that one-time screen policies would improve life expectancy by approximately 0.10 years. With Cameroon's GDP per capita being estimated at USD 1,019 
, we further estimated that a one-time screening strategy will need to cost at most 101 USD per screen for it to be considered cost-effective. This cost estimates includes both direct costs (such as cost of the test) and indirect costs (such as costs of running the screening program). More formal cost-effectiveness analyses including appropriate weighting for disability and discounting will be needed to confirm this.
As with every simulation exercise, the model form, parameters and assumptions present potential limitations to the confidence with which inferences from the analysis can be made to real-life scenarios. We choose a model that had previously been validated in the immediate pre-HAART era, a population similar to our target in this analysis. However some of the model parameters which could not be directly estimated were estimated indirectly using either US or WHO data. In the absence of prospective data on cervical cancer progression and regression rates for Cameroonian women, we used published data from the pre-HAART era in the US and assessed the impact of this choice in sensitivity analyses. A faster progression rate in developing countries could translate to even higher mortality due to cervical cancer. Nevertheless, the relative effects of HAART and screening were projected to remain similar, while the number of deaths averted with screening would increase as progression rates increase. On the other hand, if HAART was shown to have an effect on the progression of lesions, a finding that has been inconsistent 
, then the mortality due to cancer as well as the gains associated with screening would be reduced.
A further limitation, by design, of this analysis is the absence of any consideration for HPV infection status and/or HPV vaccines. The former was due to the absence of reliable data on HPV and the progression of lesions in HIV in sub-Saharan Africa. It is thus difficult to infer on the potential impact of an HPV test. We however suspect that one-time screening with an HPV-test could be better than a one-time cytology screen, as the HPV test will have a higher sensitivity, be able to detect lesions at earlier stages (allowing for early treatment) and thus result in lower mortality. Indeed, a recent study in population (of mainly HIV-negative women) in rural India reported a 50% reduction in cervical cancer mortality (measured as hazards of mortality) following a one-time HPV test 
. They also report a low insignificant reduction in mortality with cytology and VIA. These findings however need to be verified in a population of HIV-positive women, in whom the greater variability of HPV types, higher HPV-persistence rates and faster progression of lesions could mean reduced effect of an HPV test used for one-time screening.
While cervical cancer screening could substantially reduce mortality due to cervical cancer, it was projected to have very little effect on all cause mortality. We did not assess the quality of life gains, but these data indicate that even with screening further gains in overall life expectancy will depend on the extent of prevention and care for other causes of mortality including opportunistic infections.
In conclusion, cervical cancer could account for a high proportion of deaths among HIV- positive women in Africa once they have access to HAART. These deaths could be reduced with screening, even when done just once. Antiretroviral treatment scale-up activities need to be followed by strategies to systematically increase access to screening services as well as treatment for cervical precancerous and cancerous lesions as needed. While the feasibility and cost-effectiveness of more frequent screening is still to be assessed, women need to be provided the opportunity to get screened at least once on initiating HAART or at age 35.