DRUG AND COMORBID TRAJECTORIES: JOINTLY-MODELED, DIFFERENTLY-DISTRIBUTED OUTCOMES

PRINCIPAL INVESTIGATOR:  MIKULICH-GILBERTSON, SUSAN KAY  

GRANT NUMBER:  R01 DA034604-02S1

ABSTRACT:  Our R01 (DA034604 Drug and Co-morbid Trajectories: Jointly-Modeled, Differently Distributed Outcomes) directly aligns with the research objectives of this PA-13-275. Because trajectories describing drug use and comorbidity over time are often nonlinear and based on non-normal outcomes, analytic methods must evolve to understand their inter-relationships. Our R01 is developing methods that use latent stochastic parameters from multivariate nonlinear mixed (MvNLMIXED) models to evaluate inter-relationships among jointly-modeled longitudinal variables with different distributions and then applying those methods to data from two clinical trials of ADHD and substance use disorders in adolescents. By effectively leveraging the R01 team and resources, supplement funds would allow (1) application of the R01 methods to data from a 3rd trial of bupropion in youths with co-occurring ADHD, nicotine and cannabis use disorders that also evaluates depression (Bup) and (2) pursuit of a simpler approach using empirical Bayes (EB) predictors of trajectories and their parameters for individuals generated from mixed models in second stage analyses to more easily assess inter-relationships among polysubstance use and comorbidity. This approach has exciting potential but we need to determine when it can appropriately be utilized. Like many studies, Bup collected reports of daily drug use via Timeline Followback (TLFB). TLFB data is typically collapsed into summaries like total days of use that are problematic and ignore a wealth of information. An inability to fully utilize non-normal longitudinal data like TLFB information prevents us from satisfactorily addressing important issues about temporal relationships between (1) use of two drugs (e.g. does cigarette use increase or decrease with decreasing marijuana use and if so, is the change simultaneous or delayed in one?) and (2) co-morbidity and drug use (e.g. does depression improve after drug use is reduced and if so, how long after?). Bup has commonalities with the data being analyzed in the parent R01 but also provides unique measures of depression and urine biomarkers for drug use that can be used to determine inter-relationships and order and timing of changes in these co-occurring disorders during treatment. S-Aim 1 will apply parent R01 methods to Bup data in order to determine: (a) how change in use of one drug relates to change in use of others, (b) how change in ADHD severity relates to change in polysubstance use, marijuana use alone, and depression, and (c) how change in depression severity relates to change in use of cigarettes, marijuana, and alcohol. S-Aim 2 will determine under which conditions an easier, more flexible method for evaluating inter-relationships using EB predictors can be appropriately utilized, thereby increasing the utility and dissemination of methodology that targets specific interests of each institute in evaluating trajectories of alcohol (NIAAA), cigarette (NCI) and other drug (NIDA) use with each other and with comorbidity (depression, ADHD). These novel methods are widely applicable to evaluating temporal relationships among any longitudinal variables, benefiting new research and enhancing the scientific value of existing databases.

 

Back to the CRAN funded Research Portfolio