Using SMART experimental design to determine more effective interventions

By Fatima Ahmed, Assistant Research Biostatistician

Consider a treatment strategy that allows for better management of a disease or disorder. For example, a patient with ADHD receives behavior therapy and shows improvement, so the clinician encourages them to continue that treatment. Had the patient shown little or no improvement, the clinician would have had to decide whether to intensify the therapy, lengthen the duration of the therapy, and/or add medication to the treatment regimen.

How can practitioners assess treatment options and decide which one would be most effective for the patient? Adaptive interventions are systematic and replicable ways of proceeding through a sequence of pre-determined rules that guide whether, how, and when to modify interventions. Also called dynamic treatment regimens or adaptive treatment strategies, this approach adapts or modifies treatments based on a patient’s ever-changing needs.

The adaptive intervention strategy has applications beyond medicine. In fact, the strategy can be helpful to a variety of fields where interventions are regularly employed, including education,  social work, public health, and other fields.  

How can researchers support practitioners in developing effective intervention protocols? Successful adaptive interventions have been tested to ensure the strength and replicability of the outcomes of the intervention.  Many researchers are implementing experimental designs that allow them to answer specific questions to optimize these adaptive interventions.

One such design is the Sequential Multiple Assignment Randomized Trial (SMART), an experimental design used to answer multiple research questions about the selection and integration of components of an adaptive intervention. According to the Data Science for Dynamic Intervention Decision Making Center, a SMART involves multiple stages of randomizations, meaning that some or all individuals participating in a SMART are randomized more than once. This is the SMART’s defining feature. With SMART design, investigators can explore a variety of questions, including:  

• Which among a set of adaptive interventions compared produces the best overall outcome?
• Which are more cost-effective than others?
• Which treatment is the best first treatment among the choices?
• When is the best time to modify treatments?
• What are the best criteria for changing treatments, e.g. response to initial treatment, problems with compliance, specific needs, etc? 
• In study populations with large variation in response to treatments, what works for one may not work for the other.  Which adaptive intervention is best for specific subpopulations? 

In Get SMART with Adaptive Interventions, Kelley Kidwell, a professor of biostatistics at the University of Michigan and a researcher with The Data Science for Dynamic Intervention Decision Making Center, will give an informative introduction to the SMART design concept, and Liliane Windsor, Associate Dean for Research at the University of Illinois School of Social Work, will discuss the application of SMART designs in research. Attendees will have an opportunity to ask questions after the presentation.