Project summary/Abstract:
Several issues in the design and analysis of clinical trials of stimulant use disorders have impeded the
development and approval of new treatment options. A key issue concerns the reliance on abstinence as the
primary outcome. Although abstinence is a clinically relevant endpoint, relatively few trial participants ever
achieve enduring abstinence either during treatment or in the 6-12 months post treatment. Moreover, the use
of a binary measure of abstinence, relative to a quantitative measure of reduction in frequency of use, entails a
significant loss of information and statistical power. One important implication of this loss is that trials with
larger sample sizes are required, creating greater challenges for both recruitment and retention. In addition, the
statistical analyses of these trials pose a number of challenging issues that, so far, have not been adequately
addressed. Primarily, missing data is a formidable problem that is often handled with stopgap methods that
yield biased estimates of treatment effects. The overarching goal of the proposed research is to establish that
within-treatment reduction in frequency of use, as determined from urine drug screens, is (i) a more sensitive
endpoint for assessing the benefits of treatments for stimulant use disorders, and (ii) predictive of longer term
post-treatment follow-up measures of drug use and functioning. In doing so, we propose a rigorous statistical
approach to the analyses of stimulant use disorder trials that overcomes the missing data challenges
highlighted above. Specifically, using five existing clinical trial datasets, the proposed research will provide both
statistical and empirical evidence that reduction in frequency of use (based on urine drug screens), rather than
abstinence, is a more sensitive endpoint for determining treatment efficacy. In addition, the research will
examine the association between within-treatment reduction in frequency of use and longer-term follow-up
improvements in measures from five problem domains that have significant societal consequence: (i) legal, (ii)
employment, (iii) family, (iv) psychological and (v) medical problems. Finally, the research will determine
whether a key set of measures from these problem domains can be selected for use in conjunction with
reduction in frequency of use to provide the most salient and clinically meaningful endpoints for clinical trials of
stimulant use disorders.
Additionally, we plan to make the statistical methodology developed for each aim widely accessible to non-
statisticians. Specifically, we will create macros and procedures which can be used with existing statistical
software packages (e.g., SAS, Stata, and R). Statistical macros and procedures will be documented and made
publicly and freely available to other researchers in the scientific community via our website, together with
documentation on how to apply these macros to the example datasets analyzed in the resulting publications.
