Sequential Multiple Assignment Randomized Trial

---

The Sequential Multiple Assignment Randomized Trial is a type of sophisticated trial design that is used to effectively construct an adaptive intervention or Dynamic Treatment Regime. As the name suggests, the SMART design involves multiple stages of randomization, where each stage corresponds to a critical clinical time point, and participants are randomly assigned to one of the available treatment options.

Background

Randomized Control Trials

Randomized Control Trial (RCT) is a type of study design that randomly assigns participants into a treatment group or a control group. The study is conducted in such a way that the only expected difference between the control and treatment groups in an RCT is due to the outcome variable being studied. RCTs are often viewed as the gold standard clinical evidence to determine the true relative efficacy of an intervention. There are several advantages to RCTs when compared with their non-randomized counterparts, chiefly their ability to reduce bias and confounding that may be associated with factors that could otherwise influence both group assignment and prognosis.

Dynamic Treatment Regimes

A treatment regime is a set of guidelines or rules that outline how to assign treatment based on an individual patient's characteristics. When treatment is given just once, it typically occurs at a single decision point, often coinciding with the time of diagnosis . However, in numerous scenarios, particularly when dealing with chronic diseases, treatments are typically administered at multiple junctures of decision-making. An example of this is the Dynamic Treatment Regime (DTR). A DTR is also known as Adaptive Intervention (AI) or Adaptive Treatment Strategies (ATS).

A DTR comprises a series of decision rules, one for each decision point or stage, outlining how to adjust the treatment type, dosage, and timing based on continuous updates of an individual patient's information. Each decision rule considers a patient's treatment history and current status of adherence to the treatment as inputs and provides a recommended treatment as output. Decision points can be scheduled at regular intervals, such as yearly follow-up visits, or triggered by clinical events like relapse or complications. The time points at which the results of the prior treatment are assessed to make decisions regarding whether to maintain or modify treatments are termed stages within a DTR. While a DTR can theoretically encompass any number of stages, two-stage DTRs are the most commonly encountered in real-world scenarios.

File:Adaptive interventions.png

For instance, the diagram illustrates a typical example of a DTR. In this scenario, the objective is to deliver treatment to children with Attention Deficit Hyperactivity Disorder (ADHD). This DTR comprises two distinct stages of treatment provision. Initially, participants receive low-dose medication as the first stage of treatment. After 8 weeks, we move on to the second stage of treatment after classifying each child as either a responder or non-responder, using medically approved indicators. If a child positively responds to this initial treatment stage, it is continued. However, if a child does not respond, they are provided Behavioral Intervention as the second stage treatment.

In a DTR, the specific sequence of treatments that each participant undergoes is referred to as the treatment path. Using the example above, there are two potential treatment sequences that a child may follow. If we label the first-stage low-dose medication as treatment A and the second-stage behavioral intervention as treatment B, these paths can be expressed as follows:

$\bullet \text{<mrow data-mjx-texclass="ORD">Path</mrow>}1:A,A\bullet \text{<mrow data-mjx-texclass="ORD">Path</mrow>}2:A,B$

Then, a DTR can be written in terms of these paths as

$DTR=\{\begin{array}{ll}\text{Path 1},& \text{if the child is a responder}\\ \text{Path 2},& \text{if the child is a non-responder}\end{array}$

In one line, this is written as $DTR=[A,A^R{B}^{(1-R)}]$. In this context, $R$ represents the response status, where it takes a value of 0 if the participant responds to the treatment and 1 if they do not.

SMART

In the development of Dynamic Treatment Regimens (DTRs), an effective approach involves utilizing a research design known as a Sequential Multiple Assignment Randomized Trial (SMART). This framework is commonly employed in clinical research to evaluate DTRs, allowing for the examination of optimal treatment sequences and personalized treatment plans, which ultimately leads to enhanced patient outcomes.

Essentially, when managing medical conditions that see improvements under the dynamic treatment approach, we typically have multiple candidate DTRs to consider for potential treatment administration. During the trial phase, our goal is to compare all these DTRs and assess which one yields the most effective results. This objective is achieved by constructing a SMART design that encompasses all candidate DTRs, followed by gathering the necessary data from the experiment to determine the best-performing DTRs. SMART designs involve randomizing patients to available treatment options at the initial stage, followed by rerandomizing some or all of the patients to treatments available at each subsequent stage. These rerandomizations and sets of treatment options depend on how well the patient responded to the previous treatments. SMART designs offer valuable insights for tailoring healthcare interventions to individual needs, thus enhancing their effectiveness and efficiency.

File:SMART diagram.png

The diagram illustrates a SMART design using the scenario of treating children with ADHD. This design incorporates two stages of randomization. Initially, at the start of a school year, the entire child population is randomized into two first-stage treatment groups: (1) Medication and (2) Behavioral therapy. After an 8-week period, their response to the treatments is assessed, categorizing them as responders or nonresponders based on specific criteria. Responders continue with their respective first-stage treatments, while nonresponders undergo a second randomization. Nonresponders are assigned to either an intensified version of their initial treatment or a combination of the two first-stage treatments. For instance, a nonresponder to behavioral therapy might be randomized into intensified behavioral therapy or a combination of behavioral therapy and medication.

Similar to a DTR, a SMART design also incorporates a number of treatment paths. In the above diagram, if we denote T1 and T2 as the two first-stage treatments, then the six treatment paths in this SMART design can be written as

$\bullet \text{<mrow data-mjx-texclass="ORD">Path</mrow>}1:T1,A\bullet \text{<mrow data-mjx-texclass="ORD">Path</mrow>}2:T1,B\bullet \text{<mrow data-mjx-texclass="ORD">Path</mrow>}3:T1,C$

$\bullet \text{<mrow data-mjx-texclass="ORD">Path</mrow>}4:T2,D\bullet \text{<mrow data-mjx-texclass="ORD">Path</mrow>}5:T2,E\bullet \text{<mrow data-mjx-texclass="ORD">Path</mrow>}6:T2,F$

These six treatment paths induce four DTRs that are essentially considered within the SMART design and are compared to each other for us to get the DTR that provides optimal results. Following similar notations in the previous section, the DTRs are given as

• $DTR1=[T1,A^R{B}^{(1-R)}]$
• $DTR2=[T1,A^R{C}^{(1-R)}]$
• $DTR3=[T2,D^R{E}^{(1-R)}]$
• $DTR4=[T2,D^R{F}^{(1-R)}]$

These four DTRs are termed as the embedded DTRs of the SMART design depicted above.

Advantages of a SMART design

Compared to a conventional controlled trial or a series of single-stage trials in the development of optimal DTRs, SMART designs present several advantages. In single-stage trials, the primary focus is often on immediate treatment responses, which may not account for delayed therapeutic effects. In contrast, SMART designs allow for the identification of such delayed effects. Moreover, SMART designs contribute to more precise diagnostic effects by guiding better treatment decisions, even in cases where the initial treatment may not be highly effective. Additionally, trials employing SMART designs tend to have improved recruitment and retention rates compared to standard trials. This is because, in single-stage trials, non-responding patients may be more likely to drop out, while in SMART designs, the probability of patient dropout diminishes since they anticipate the possibility of receiving improved treatments in subsequent stages.

See also

• YouTube video about SMART from Johns Hopkins University Alacrity Center for Health and Longevity in Mental Illness.
• YouTube video (Part 1) and YouTube video (Part 2) about Dynamic Treatment Regimes and SMART Design.
• Webpage for Adaptive Intervention and Webpage for SMART from the Data Science for Dynamic Intervention Decision Making Center.

References

This article "Sequential Multiple Assignment Randomized Trial" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:Sequential Multiple Assignment Randomized Trial. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.

---