1	Clinical Trial Design Issues in Development of Topical Microbicides Tim Farley Department of Reproductive Health and Research World Health Organization Geneva, Switzerland
2	Overview Measures of product effect Efficacy, effectiveness and use-effectiveness Choice of control arms Placebo product or no product (condom only) Strength of evidence Are two independent trials at P < 0.05 ethical?
3	Efficacy vs. Effectiveness Efficacy prevented fraction of infections when product used In practice, product not used consistently observe effectiveness = prevented fraction under conditions of typical use Despite collecting information on when product used and when not used, cannot estimate efficacy from microbicide trial
4	Use-effectiveness Use in general population, beyond carefully controlled and artificial research setting less well-trained or -informed users changes in sexual behaviour and condom use inconsistent supply, … May be very difficult to estimate Requires good measure of incidence in target population when product not available or before introduction Directly comparable information lacking
5	Choice of Control Arm Randomization Ensures balance of factors related to individual risk and to patterns of condom and product use Cannot balance changes of behaviour once study group has been revealed Require good masking (or blinding) Placebo-controlled double-blind trial Preferred whenever feasible Gives unbiased estimate of product effectiveness
6	No-product Arm? Essential when no placebo product available Cannot rely on randomization and blinding to balance behaviours and condom use Must collect high-quality, extensive and reliable data on product and condom use Analysis adjusted for reported behaviours Expected misclassification dilutes estimated effect Two control groups? Costly, no benefit for interpretation, potentially confusing, (e.g. COL-1492)
7	Strength of Evidence Two independent studies at P < 0.05 Desirable Ethical Review Committees unlikely to approve Single study at P < 0.0013 equivalent to two independent P < 0.05 studies Also unethical Single P < 0.05 study may not convince When would a second study be no longer ethical? P < 0.05, 0.04, 0.03, 0.02, 0.01, …?