Presentations and Workshops on Adaptive Interventions and Sequential Multiple Assignment Randomized Trials

[ Daniel Almirall's Home Page ][ Billie Nahum-Shani's Home Page ][ Susan Murphy's Home Page ][ Examples of SMART Studies ]

Workshop on Adaptive Interventions and Sequential Multiple Assignment Randomized Trial (SMART) Designs

This section includes slides (and sample SAS code) for six modules we use in a typical 2-day workshop on adaptive interventions and the design and analysis of Sequential Multiple Assignment Randomized Trials (SMART). These modules were developed in collaboration with my colleagues, Billie Nahum-Shani and Susan A. Murphy; all of us contributed equally to this work. Audrey Boruvka translated to R all of the SAS code used to illustrate the methodology (see below). Support for this work was provided by the NIH/NIDA, via a grant to the Methodology Center at Penn State University (P50-DA-010075).

Workshop Schedule / Detailed Outline
Workshop Agenda	This is a detailed outline of Days 1 and 2 of a typical 2-day workshop, including a description of modules and exercises.
Workshop Preparation
Workshop Preparation Instructions	These are pre-workshop instructions. Follow these instructions to install PROC QLEARN and to test that SAS works on your machine. We will use SAS on Day 2 of the workshop. PROC QLEARN is a SAS add-on that implements Q-Learning Regression.
Workshop Preparation Files	You will need the files contained inside of this ZIP to go through the Workshop Preparation Instructions. To download, right-click and select "Save Link As..."
Slides and Handouts
Module 1 Slides, AIs	Adaptive Interventions (AIs): What are AIs? Why do we need them? How can AIs be used to inform clinical practice?
Module 2 Slides, SMART	Sequential Multiple Assignment Randomized Trials (SMART): What are SMARTs? Why do we need them? SMART design principles. SMART sample size considerations.
SMART Case Studies Handout	SMART Case Studies Handout: A PDF file describing the five (5) SMART to be discussed in Module 3 on SMART Case Studies.
Module 3 Slides, SMART Case Studies	A review of five (5) actual SMARTs. These slides compare and contrast each of the four SMART study designs on dimensions such as: rationale for the SMART, what are the SMART-embedded tailoring variables, what are the SMART-embedded adaptive treatment strategies, what are the primary analyses, what are the secondary analyses.
ADHD Handout	This handout describes the ADHD SMART study (PI: William Pelham, Florida International University) and the variables in a data set that was simulated to mimic the ADHD SMART study (see below for data set). The ADHD SMART study is used as the prototypical SMART design to describe the SAS analyses in Modules 4, 5 and 6.
Module 4 Slides, Primary Analyses I	Primary Data Analyses using data arising from a SMART, Part I: Main effect of first-stage tactics. Comparing second-stage tactics. We use the ADHD SMART study as an example in these slides.
Module 5 Slides, Primary Analyses II	Primary Data Analyses using data arising from a SMART, Part II: Weighting-and-replication approach to estimate and compare mean outcomes under the SMART-embedded adaptive treatment strategies. We use the ADHD SMART study as an example in these slides.
Module 6 Slides, Secondary Analyses	Secondary Data Analyses using data arising from a SMART: Q-Learning Regression to develop more richly-individualized adaptive treatment strategies using data arising from a SMART. This module also teaches investigators how to use SAS add-on PROC QLEARN to implement Q-Learning Regression using data arising from a SMART. We use the ADHD SMART study as an example in these slides.
SAS Code to Accompany Slides and Practice Exercises
ADHD Handout	This handout describes the ADHD SMART study (PI: William Pelham, Florida International University) and the variables in a data set that was simulated to mimic the ADHD SMART study (see below for data set). The ADHD SMART study is used as the prototypical SMART design to describe the SAS analyses in Modules 4, 5 and 6.
Simulated ADHD Data Set	The ADHD SMART study has been completed. This data set was simulated to mimic data arising from the ADHD SMART study. This is NOT the actual ADHD SMART study data set. [SAS data format][TXT data format]
SAS Code to Analyze ADHD data	This file provides SAS code to implement the analyses in Modules 4, 5, and 6 using the simulated ADHD data set.
SAS Output for ADHD Data Analysis	After running the SAS code above, your SAS output should match the contents of this file.
R Code to Analyze ADHD data	This file provides R code to implement the analyses in Modules 4, 5, and 6 using the simulated ADHD data set. This code is provided courtesy of Audrey Boruvka, PhD.
R Output for ADHD Data Analysis	After running the R code above, your R output should match the contents of this file. This output file is provided courtesy of Audrey Boruvka, PhD.
Other Workshop Files
FAQ	We have begun compiling a Frequently Asked Questions handout based on common questions and concerns that are raised about SMARTs during our workshops. Please email Daniel Almirall with questions or concerns to add to this list.

First Published: 4/29/2011; Last Revised: 07/16/2014