Beyond the trials Translating research results into public
Beyond the trials: Translating research results into public health impact (an update on modelling) Catherine Hankins MD MSc FRCPC Chief Scientific Adviser to UNAIDS Office of the Deputy Executive Director The Promise and Perils of antiretroviral-based prevention: Making it a reality on the ground Satellite Session at AIDS 2010 Sunday, 18 July 2010
Beyond the trials: Translating research results into public health impact (an update on modelling) • What modelling can contribute to decision-making • Modelling approaches • Modelling and stakeholder deliberations • Initial findings • Questions from modellers
What modelling can contribute to decision -making • • Explicit assumptions – testable predictions Framework for data analysis Projections From outcome to impact (effect for costeffectiveness analysis) • Assessing: – Perverse outcomes – Combining interventions – Impact of new technologies • Setting coverage targets • Advocacy
What works for HIV prevention: Results from randomised controlled trials (RCTs) with HIV incidence end points
Key Equations ra, s, t = ru x Avg. Riska, s, t x BCa, s, t x MCeffecta, s, t x Prev. Effectt ru = r’/[(1 -MC%) + MC% x (1 -Red)] r’ = initial force of infection, from curve fit Avg. Risk = e-αI/N reduction in average risk among susceptible population as highest risk populations become infected BC = e-εD behavior change (D = Cumulative AIDS deaths) MCeffect = (1 -MC%t) + ∑p(Ppx. MC%x. Redp)/∑p. P Prev. Effect = future scale-up of prevention coverage
Discounted Costs and Savings
Male circumcision modelling approaches and settings P. I. Williams Model type Closedsolution analysis & deterministic Setting Africa & South Africa Nagelkerke Closedsolution analysis & deterministic Gray Stochastic Botswana Rural and Nyanza, Uganda Kenya Hallett White Alsallaq Determin- Stochastic network istic Closedsolution analysis & determin -istic Southern /Eastern Africa Kisumu, Kenya & Africa simulation Kisumu, Kenya
Expert Review Group Consensus
Population-level Impacts by Coverage
Male Circumcision: Moving from Research to Implementation
Importance of epidemiological context in deciding who should be prioritised for Pr. EP
Pr. EP Modelling Overview Existing modelling work is limited but rapidly expanding: – General population, heterosexual transmission in sub. Saharan Africa Abbas et. al. , 2007 – General population, sex workers & their clients in Africa and India Vissers et. al. , 2008 – Men who have sex with men in New York City Desai et. al. , 2008 – Men who have sex with men USA Patiel et. al. 2009 – Impact of resistance on Pr. EP effectiveness in Zimbabwe Van de Vijver et. al. , 2009
Increases in sexual risk behaviour may erode or reverse preventive benefit of Pr. EP Abbas et. al. Note different scales for increased risk Desai et. al. Copyright © 2010 AIDS. Published by Lippincott Williams & Wilkins. 13
Communicating about partial protection
Pr. EP Modelling Overview: Preliminary Findings Considerable potential preventive benefit if Pr. EP: • shows high effectiveness • is prioritised for those at highest risk of HIV exposure • does not lead to risk compensation (increased partner numbers, decreased condom use, choice of more risky partners) • adherence is high Circulating drug resistance would have a limited impact on Pr. EP effectiveness Risk of resistance is highest when Pr. EP is used by someone who has HIV infection already
Pr. EP Modelling 2010 -2011 • UNAIDS/WHO Pr. EP Modelling Meeting, Geneva March 2010 (consensus meeting planned for 2011) • Georgetown/Imperial/WHO/UNAIDS supported by BMGF to convene 5 regional consultations for which Imperial is preparing region-specific modelling scenarios: – – – West Africa East Africa Southern Africa Latin America Asia
Pr. EP in Combination Prevention Status quo Intervention to scale (incr. condom use and prompt treatment initiation) + Targeted effective Pre. P + The missing piece? Numbers based on extrapolation to Urban Benin; *Pre. P intervention is to 60% of sex workers & clients; 70% efficacy and 80% adherence, for 10 years. ** The missing piece required to reduce incidence by 90% in 2031 and eventually stop the epidemic is a 60% efficacy vaccine delivered to half the population.
Other contexts: Preliminary Results Hyper-endemic setting IDU and MSM
Pr. EP Serodiscordant couples (UNAIDS/WHO Modelling Group and Imperial College)
Questions from modellers • What efficacy range will be acceptable: what would be worth implementing and where? • What range of HIV testing frequency? – Frequency will determine length of time a person with acute infection will be on suboptimal therapy • Implications of development of resistance on the cost of eventual treatment – use of a non. TDF first line regimen • Implications for transmissibility during primary infection
Questions from modellers • Pending bridging studies in pregnant women, would delivery be for men only? Should indirect effects for women be modelled? • What level of adherence likely in which populations? • What discontinuation rate and why (toxicity vs. other reasons) • Potential for risk compensation (behavioural risk enhancement)? – Defined how? (increased frequency, number of partners, risk of partners, unprotected sex acts)
Questions from modellers • Heterogeneity in susceptibility to infection by host genetics, mode of transmission, founder virus/es • Assumptions on the extent of scale-up of other interventions when Pr. EP introduced and potential synergies or displacement of resources/attention: MC, ART, condom promotion • Policy re discordant couples: treat the HIV-positive person and provide Pr. EP one year to the HIVnegative person? • Time to achieve target coverage? (S-shaped, linear) • Service delivery: cadre of health worker, site, SMS texting….
Acknowledgements • • • John Stover Geoff Garnett Tim Hallett Ide Cremin Brian Williams Lori Bollinger Salim Abdool Karim Quarraisha Abdool Karim Toby Kasper
Thank you for your attention
- Slides: 25