W-26: Statistical Issues in Human Drug Abuse Study – Two Different Types of Studies
Lead Statistician, OB, OTS, CDER
FDA United States
To illustrate the procedure and statistical methodologies in human drug abuse study, including human drug abuse potential study (HAP) and human abuse study for abuse-deterrent formulation (ADF) opioids
We will discuss implementation on issues such as study design, sample size calculation, hypothesis testing, testing margins as well as modeling with hypothetical and real examples. We will show examples on different scenarios and the optimal options.
HAP studies in both applications are very similar but intrinsically different in statistical approach as one is an efficacy studies for their abuse-deterrent properties while the other belongs to safety studies. The test margin has to be pre-determined but ad-hoc approach can be applied to obtain the actual margin in data analysis for ADF drugs. On the other hand, test margins in general HAP studies are related to the choice of positive controls. The test procedure needs to be a closed procedure to control for type I error rates. Varies types of t-test and non-parametric approaches can be considered when normality test failed, but the results should be carefully interpreted as we switch to non-parametric models.
Human drug abuse study is an important component in evaluating whether or the magnitude of a drug produces subjective responses related to abuse potential through the drug effects on the human central nerve system (CNS). Correct approach on both study design and data analysis ensures the relative abuse potential of a test drug (NME or ADF) is adequately evaluated through subjective (VAS score for drug liking, take drug again, etc.) and objective measures (pupil dilation, etc.) on recreational drug users.