Accelerating clinical development of HIV vaccine strategies: methodological challenges and considerations in constructing an optimised multi-arm phase I/II trial design

Table 1 Comparison of operating characteristics of the two types of safety decision rules

Type of safety decision rule		Simulation scenario
	Vaccine safe (P_safe = 0.95)	Vaccine unsafe (P_safe = 0.7)
	Probability of stopping	Probability of stopping	Participant number at stop (median; IQR)
Fixed-sample frequentist rule at n = 19	0.07	0.95	19	(19-19)^a
Continuous Bayesian monitoring	0.05	0.96	8	(4-12)

Fixed-sample frequentist rule: Interim analysis after 19 participants. Vaccine considered unsafe at interim if lower bound of one-sided exact 95% confidence interval of observed proportion ≤0.7 for safety endpoint.
^aIn a strict application of the fixed sample-design the decision rule only applies when the outcome of 19 participants has been observed. If the application of the rule was handled more flexibly (that is stopping as soon as a third participant experiences an vaccine-related grade 3 or 4 adverse event, regardless of the denominator), stopping would occur after nine participants in median (IQR 6-12).
Continuous Bayesian monitoring of the safety endpoint: Sequential analysis after each participant or event. Vaccine considered unsafe at the first analysis where the posterior probability that the vaccine is below the target level (Psafe <0.95) exceeds 95%. Enthusiastic prior: beta (6,0.3).

ISSN: 1745-6215