Statistical considerations for a phase 0 trial

Abstract

14038 Background: Phase ‘0’ clinical trials are small first in man studies to determine a dose or dose range which results in a desired biologic effect. Novel statistical designs are needed due to the limited number of patients and the complexities of analyzing primary pharmacodynamic endpoints. Methods: In general, the effect of the agent will be evaluated, based on a biologic endpoint(s), at two levels, the individual participant level (for a given dose amount), and the agent dose amount level (across individual participants). At the participant level, the effect is preferably defined as a dichotomous outcome. The statistical criteria should be such that the threshold is not surpassed with more than .10 probability, for a given participant, at a given dose amount, for the null hypothesis case of no true biologic effect. At the dose amount level, the effect rate is defined as the probability that a randomly chosen participant will satisfy the participant level threshold. The statistical criteria for declaring an observed effect rate to be significant should be such that the over-all probability of declaring the agent effective (summed across all dose amounts) should not exceed .10 for the null hypothesis case of no true biologic effect. A target effect rate, measured across participants, should be given, and the power to detect it, for a given dose amount, should be estimated. Ideally, there will be multiple pre-agent administration endpoint measurements, per participant, to define variability. Examples: 1. If 10 participants are accrued to a single dose amount, an observed effect rate for that dose amount could be defined to be significant if at least 2 of the 10 participants satisfy the participant threshold, yielding 85–90% power to detect a 35% effect rate. 2. If 5 participants are accrued to each of 2 dose amounts, an observed effect rate for a particular dose amount could be defined to be significant if at least 2 of the 5 participants satisfy the participant threshold, yielding approximately 90% power to detect a 60% effect rate. 3. If 3 participants are accrued to each of 4 dose amounts, an observed effect rate for a particular dose amount could be defined to be significant if at least 2 of the 3 participants satisfy the participant threshold, yielding approximately 90% power to detect an 80% effect rate. No significant financial relationships to disclose.