1	Statistical Considerations for Topical Microbicide Phase 2 and 3 Trial Designs: A Regulatory Perspective Rafia Bhore, Ph.D. Greg Soon, Ph.D. Division of Antiviral Drug Products Food and Drug Administration
2	Outline Phase 2/3 Clinical Trial Design: An Example Two Arms or Three Arms ? p-value (Single Trial vs. Two Trials) Criteria for a “Win”/Success Power Considerations Sample Size Estimates Other Considerations
3	Phase 2/3 Clinical Trial Design An Example
4	"To establish safety and efficacy" To establish safety and efficacy of investigational microbicide in preventing HIV infection Test Group Control Group 1 Control Group 2 Microbicide “Placebo” + + Condom Condom Condom-only (“no-treatment”)
5	Two Arms or Three Arms?
6	“Placebo” Arm “Placebo” provides a means to blinding investigators and participants Maximizes the likelihood of obtaining an unbiased estimate of efficacy Can we assume “Placebo” is inert? Antimicrobial activity of “Placebo” not known/not proven. Protective effect or harmful effect?
7	Condom-only (“No-Treatment”) Arm Established standard for prevention of HIV Provide comparison of “real world” effectiveness in preventing transmission Data on sexual behaviors associated with use or non-use of microbicide products Single component of other arms containing gel + condom
8	P-value (Single vs. Two Trials) (Significance Level = Probability of Type 1 error: false positive)
9	Level of Evidence
10	P-value (Two trials) Two trials (Regulatory Standards) Trial 1 p-value < 0.05 (two-sided) Trial 2 p-value < 0.05 (two-sided) Run in parallel or staggered in time? If staggered, how much gap in time?
11	P-value (One trial) Need to show as strong and robust evidence as two separate trials May not be repeatable ethical concerns ? One trial: p-value < 0.001 (two-sided) (because 2x[0.025^2]=0.00125) same as p-value < 0.0005 (one-sided)
12	Replicating a Study Result Probability of observing a statistically significant result (e.g. p < 0.05) upon repetition of a clinical trial when the effect size observed in the first trial is assumed to be the true effect
13	Overall Level of Evidence p-value < 0.001 considered convincing p-value >= 0.01 would be inadequate p-value between 0.001 and 0.01 possibly adequate, provided that results are consistent across various subgroups other supportive evidence is strong
14	Criteria for a “Win”/Success
15	Definition of a “Win” HIV infection rate in Microbicide < “Placebo” p-value < 0.001 (two-sided) AND Microbicide < Condom-only p-value < 0.001 (two-sided)
16	Why win versus “Placebo” arm? If the HIV infection rate in Microbicide + Condom » “Placebo”+Cond. Microbicide + Condom < Condom then is “Placebo” as effective as Microbicide? (does not prove efficacy of microbicide)
17	Why win versus Condom-only? If the HIV infection rate in Microbicide + Condom < “Placebo” + Cond. Microbicide + Condom » Condom then the use of microbicide in conjunction with condom does not provide any additional protection than condom alone
18	Definition of a “Win” HIV infection rate in Microbicide < “Placebo” p-value < 0.001 (two-sided) AND Microbicide < Condom-only p-value < 0.001 (two-sided)
19	Sample Size Estimates
20	Factors Affecting Sample Size HIV Seroincidence Rates in Control 0.5, 6, 7, 9 per 100 person-years Effect Size (33%, 50%, … ) Length of Follow-up 12 / 24 months for each participant Last patient completes 12 / 24 months Statistical Power (90%, 80%, …)
21	Sample Size Estimates (Duration of study=24 months, Power=90%, a=.001)
22	Power Considerations (Power = 1-Probability of Type II error [false negative])
23	Example Test two sets of hypotheses H₀: Microbicide+condom = “Placebo”+condom H_A: Microbicide+condom < “Placebo”+condom H₀: Microbicide+condom = Condom-only H_A: Microbicide+condom < Condom-only Calculate sample-size Power for each test is 90% at 33% reduction in HIV infection rate from condom-only arm
24	Power Curve: Varying Rates of Risk Reduction from Placebo
25	Other Considerations
26	Follow-up Continue study until last subject enrolled completes at least 12 months on study Pro-active in following participants actively pursue and identify reasons for dropouts (safety issues?) continue follow-up after study drug discontinuation
27	Condom + Microbicide Use Monitoring Collect data on use of condom and other barrier/drug use Efficacy evidence closely tied with compliance “Frequently” collect information on number of sexual acts
28	Allocation Ratio Microbicide : Placebo : Condom-only Standard practice is to randomize 1:1:1 Alternatively randomize x: y: z e.g., 3:2:2 or 1.5:1:1 More participants in microbicide arm than control groups. Same number in both control groups Alternative allocation ratios will optimize overall power to detect a difference and increase safety data on microbicide arm.
29	Summary Three-arm design will ensure that the first generation of microbicides is appropriately studied Single trial to show same level of evidence as two separate trials Sample size depends on assumptions Length of follow-up important in observing HIV endpoints.
30	Acknowledgments Dr. Teresa Wu, Medical Officer Dr. Debra Birnkrant, Director Division of Antiviral Drug Products, FDA