PubH 7470: STATISTICS FOR TRANSLATIONAL & CLINICAL RESEARCH

SMART Trial Designs

SMART is S.M.A.R.T.

• S – Sequential • M – Multiple • A – Assignment • R – Randomized • T – Trial

For many individuals, substance abuse possesses characteristics of chronic disorders in that individuals experience repeated cycles of cessation and relapse. Viewing drug dependence as an “chronic”, relapsing disorder is increasingly accepted and effective treatment strategies are desirable for managing the variable course disorder.

There are circumstances, substance abuse and other chronic diseases included, where: (1) There are more than one treatments which

might work but none stands out nor dominating, (2) Some subjects might be successfully treated

once but other subjects might need more than one regiment; successful ones might relapse.

Therefore, strategies are needed to individualize treatment via decision rules that recommend when and how treatment should be change. Recommendations are based on patient characteristics and outcomes collected during treatment such as patient response or progress and adherence.

The development of adaptive treatment strategies requires consideration of the ordering of treatments, the timing of changes in treatment, and the use of measures of response, burden, and adherence collected during treatment to make further treatment decisions. The sequential multiple assignment randomized trial (SMART) is a newer such adaptive treatment strategy. It retains the most basic element of conventional experiment design: the randomization. Subjects are often randomized multiple times.

In addition to substance abuse, a number of SMART trials have also been conducted to deal with Depression (Lavori et al. 2001, Rush et al. 2003) , Alzheimer (Schneider 2001), Cancer (melanoma, Freda et al. 2009), and Autism (Kasari et al. 2014). As an illustration, we start with an ongoing trial for Smoking Cessation.

Clinical trials have demonstrated the efficacy of behavioral, pharmacological, and combination treatment strategies for smoking cessation; but researchers have focused on evaluating “one-time” treatment strategies delivered in isolation. Results? These treatments yield low rates of long-term abstinence – in the range of 5%-20% depending on the population of smokers and the type and intensity of treatment.

That leaves practitioners with frustration od scant guidance on how to manage smoking cessation treatment over time, especially on how to “tailor” therapy based on patients’ response to initial treatment which maybe different from patient to patient. An “adaptive intervention” is desirable; and SMART was considered. The goal is to prove that chronic care is more effective than episodic care.

Setting: Investigators recruited subjects from two Lung Cancer Screening Programs, one at the University of Minnesota and one at the Minneapolis VA Medical Center. These are older (aged 55-79) current daily smokers with a smoking history of 30pack-years or greater who were willing to choose a quit day within the next 12 months. Some exclusion criteria applied; for example, psychotic disorders or depression.

First Randomization: All participants will receive a first-line treatment starting with the quit date. First randomization will occur at baseline (“R1”, into group A or B) when participant selects a quit date; however, the participant and the counselor will be blinded to treatment assignment until 4 weeks later. First-line treatment for Group A will last 4 weeks; at this time, depending on the result of the first-line treatment, subjects will receive second randomization. First-line treatment for Group B will last 8 weeks; after that, subjects will receive their second randomization, also depending on the result of the first-line treatment.

First-line Treatment: This is the conventional one-time treatment for all participants. It consists of 4 or 8 weeks of counseling with nicotine replacement therapy (NRT). NRT maybe monotherapy (patch, gum, lozenge) or combination therapy (patch + gum, patch + lozenge). Subjects receive periodic calls; call contents include problem-solving, skills training, and social supports. Results are classified as complete responders (success) or incomplete responders (failure; any smoking after quit date).

Subsequent Treatments: After the first-line treatment, subsequent treatments include Tobacco Longitudinal Care (TLC) and Medication Therapy Management (MTM), an rather intensive pharmacological treatment. The total intervention program will be 12 months regardless of randomized treatment group and regardless of transitional outcomes.

Tobacco Longitudinal Care (TLC) is a well-established care model for tobacco treatment, including recycling treatment tools and relapse prevention. Chronic disease management includes continuity of care, stepped care improvement of symptoms and self-management. The TLC protocol includes options for whether participants are abstinent, have relapsed, have resumed the original amount of smoking, or have not quit at all. It is a 12-month program; there are 2 more intense version for which calls are made every 3 months (TLC-Quarterly) or one month (TLC-Monthly)

Medication Therapy Management (MTM) is an intensive pharmacological treatment. Its services expands the toolbox available in TLC to include in-person consultation with a pharmacist, the prescription drugs Bupropion or Varenicline, or combination medications (NRT + Bupropion).

THE DESIGN

Primary Outcome: The primary (binary) outcome for all analyses will be whether or not smokers achieve 6-month prolonged abstinence measured at 18 months after the baseline assessment. All analyses will be governed by the “Intent-to-Treat” principle.

Primary Specific Aim: Among incomplete responders, long-term abstinence rates will be higher in smokers randomized to TLC + MTM compared to TLC.

Secondary Specific Aim 1: Among complete responders, long-term abstinence rates will be higher in smokers randomized to TLC compared to TLC-Quarterly Secondary Specific Aim 2A: Incomplete responders to first-line treatment who report more days smoking following initial quit date will benefit more from TLC + MTM compared to TLC. Secondary Specific Aim 2B: Complete responders identified at week 4 will benefit more from TLC compared to TLC-Quarterly, but there will be no difference between TLC and TLC-Quarterly for complete responders identified at week 8.

Statistical Analysis: The proposed trial design of “adaptive interventions” is equivalent to a factorial design. No special methods are needed except for special care to deal with: (1) Missing Data, and (2) Sample Size Determination.

RATIONALE FOR STUDY DESIGN

Further pharmacological treatment, in this proposal represented by MTM, requires pharmacy personnel with specialized clinical skills and training, face-to-face contact, and additional patient burden and expense. The Primary Aim will answer the important question of whether the additional benefit provided by a more complex clinical treatment model is sufficient to offer it on the large scale afforded by attendance at lung cancer screening programs.

RATIONALE FOR STUDY DESIGN The two supplementary randomizations embedded in our study capitalize on SMART design opportunities to examine additional questions: (1) Is a less resource intensive follow-up protocol for smokers who respond completely to first-line treatment just as good or less effective? (Secondary Aim 1) and (2) Does the timing of assessment of first-line treatment response moderate the effectiveness of the treatment subsequently provided? (Secondary Aim 2). Random assignment of treatment conditions will make these analyses more robust than the observational analyses possible in a standard randomized controlled trial.

Summary: (Conventional) Clinical Trials evaluate (one-time) treatments whereas SMART’s evaluate “treatment programs” which, in many cases, would be more useful for real life practices. The two key words are “sequential” and “multiple”.

Summary: In this particular case of a study to form a smoking cassation program, results of this SMART would help to form guidance covering: (1) optimal length of the first-line treatment, and (2) optimal choice of a subsequent treatment depending on the outcome of the first-line treatment appropriate foe each subject.

Major Goal of SMART The main goal of SMART Design is to answer scientific questions holistically and rigorously (i.e. randomized): (1) SMART trials typically consist of two phases; treatment assigned on phase II depends on the result of Phase I by separate randomizations. (2) SMART trials focus on the order and the timing in which treatments are administered to individuals in the trial.

Goals & Characteristics of SMART

• Addresses sequencing and timing questions that arise

• Used for developing new treatment strategies or refining existing adaptive treatments

• Used for time-varying adaptive interventions • Utilizes randomized experimental design • Identifies the best decision rules based on an

individual’s response to the treatment • Resulting program from a SMART trial could

be further evaluated in a conventional randomized control trial.

Adaptive Interventions To develop an adaptive intervention, specific questions must be asked:

– What is the best sequence of treatments/interventions?

– How often should the intervention be assessed as well as provide an opportunity to change the initial dose/length of treatment?

– Should treatments be assigned to individuals or should they be allowed to choose a treatment option?

Advantages of a SMART Trial • Avoids deleterious effects of a cohort study

– SMART trial provides treatment for participants who did not improve from Phase I treatment compared to simply receiving a placebo treatment

• Phase I treatment allows patient to benefit more from following treatments (i.e. Phase II treatments) – The follow-up treatment in Phase II should

provide insight into the usefulness of the treatment provided in Phase I

Advantages of a SMART Trial • SMART trial improves the chances of combined

effect through treatment sequences and timing decisions (result is a “treatment program”) – Allowing each treatment to have an adequate

amount of time to work – Provides opportunity to see how two

treatments work together to produce a favorable result

– This is done in hopes to avoid unintentional negative interactions from the treatment

Advantages of a SMART Trial • “By proactively and experimentally

investigating which of the treatment sequences and timing decisions are important, which treatment decisions interact and which treatment decisions interact with patient characteristics and outcomes, we can increase the chance that the developed adaptive treatment strategy will exhibit improved performance when evaluated in a future randomized trial” (Murphy et al, 2007).

Disadvantages of a SMART Trial

• It is a new clinical approach, acceptance still not universal; only applicable with/to some diseases

• More complex due to all of the treatments that are administered

• Subjects might withdraw consent for subsequent randomizations

• Loss to follow-up; have to deal with missing data

• Issues with treatment implementation

Some Other Examples of SMART Trials

• The CATIE Trial, a SMART Trial for antipsychotic medications in patients with schizophrenia

• STAR*D Trial, a SMART Trial for the treatment of depression

• SMART trial for the malignant melanoma patients being treated with interferon alpha therapy

• SMART trial focusing on communication interventions for children with autism

Example #2: MELANOMA • Background: Interferon-alpha therapy can

cause neurobehavioral symptoms for adults with a malignant melanoma

• Study Group: Adults with malignant melanoma who plan to receive interferon-alpha treatment (IFN)

• Goal: To determine which allocation of treatment results in the greatest adherence

• Study: Double-blind, randomized trial

Example #2: MELANOMA • Participants are classified as either a

“remitter” or a “non-remitter” before receiving Phase I treatment – Remitter – Experience a remission of

neurobehavioral symptoms – Non-remitter – Do not experience a

remission of neurobehavioral symptoms • After Phase I is complete, the researchers

find out who was a remitter or a non-remitter

MELANOMA SMART Trial • For adults classified as “remitters”:

– Phase I Intervention: Escitalopram or MPH – Phase II Intervention: Remain on initially

prescribed medication (either Escitalopram or MPH)

• For adults classified as “non-remitters”: – Phase I Intervention: Escitalopram or MPH – Phase II Intervention: Prescribed either

Escitalopram, MPH, or Escitalopram + MPH

MELANOMA SMART Trial • For both Phase I and Phase II, participants undergo

treatment for 6- 8 weeks • Phase II

– Remitter patients receive the same treatment as was allocated in Phase I

– Non-remitter patients who received Escitalopram during Phase I, will be randomly allocated to either MPH (Switch) or MPH + Escitalopram (Augment)

– Non-remitter patients who received MPH during Phase I, will be randomly allocated to either Escitalopram (Switch) or MPH + Escitalopram (Augment)

MELANOMA SMART Trial

Freda Auyeung et al., 2009

MELANOMA SMART Trial • Primary Endpoint: Which of the four

treatment strategies (as seen in Table 2) has the highest adherence rate to the interferon-alpha treatment

• Secondary Endpoint: Whether the switch or augment is a more preferable choice for “non-remitters” – Hypothesized that the augment option

would be optimal for improving treatment adherence

Example #3: AUTISM

• Communication Interventions for Minimally Verbal Children With Autism: A Sequential Multiple Assignment Randomized Trial (SMART) – Connie Kasari, PhD, et al. (2014)

• Objective: “Test the effect of beginning treatment with a speech-generating device in the context of a blended, adaptive treatment design for improving spontaneous, communicative utterances in school-aged children with autism.”

Autism SMART Trial: Background • 25%-30% of children with autism remain

minimally verbal after years of interventions • Poor social and adaptive functioning for

children that fail to develop spoken language by age 5

• Limited opportunity for children to learn spoken language after 5 years of age

• Research conducted on speech that was communicative (words directed to others) within behavioral intervention studies is limited

Terms Used • SGDs: speech-generating

device • SCUs: spontaneous,

communicative utterances • JASPER: Joint Attention Symbolic Play

Engagement and Regulation • EMT: Enhanced Milieu Teaching • TSCU: total number of SCUs which included

comments, requests, and protests • TDWR: total number of different word roots • TCOM: total number of comments

Autism SMART Trial: Design • Longitudinal study with outcome measures

at baseline, weeks 12, 24, and 36 (follow-up) • Stage 1 Treatment (12 weeks)

– JASP +EMT – JASP + EMT + SGD

• Stage 2 Treatment (12 weeks) – Treatments adapted based on how

children responded to Stage 1 treatment

Autism SMART Trial: Participants • 61 children between 5 and 8 years • Clinical diagnosis of autism spectrum

disorders (ASD) • Evidence of minimally verbal communication

(fewer than 20 words used during 20-minute baseline test)

• At least 2 years of prior intervention • Excluded: other major medical conditions,

sensory disabilities, motor disabilities, uncontrolled seizure disorders, or proficient use of an SGD

Autism SMART Trial: Aims • Primary Aim: examine the effect of the adaptive

intervention beginning with JASP + EMT + SGD vs. beginning with JASP + EMT

• Secondary Aim: compare outcomes across 3 adaptive interventions – First adaptive intervention: JASP + EMT

intensified JASP + EMT – Second adaptive intervention: JASP + EMT

augmented JASP + EMT with SGD – Third adaptive intervention: JASP + EMT + SGD

intensified JASP + EMT+ SGD • Next phase in each intervention was for slow

responders, otherwise intervention maintained

Autism SMART Trial: Measures

• Natural Language Sample (NLS): used to evaluate a child’s spontaneous expressive language ability – TSCU, TDWR, and TCOM outcomes – Spoken and SGD-produced utterances

used – Collected at baseline before treatment, and

at weeks 12, 24, and 36 (follow-up) • Many other assessments used to confirm

baseline eligibility for ASD and IQ scores

AUTISM: Stage 1 Treatments • Weeks 1 to 12 • Two 1-hour sessions per week • Participants randomized to:

– JASP + EMT – JASP + EMT + SGD

• SDG used about 50% of the time by interventionist when modeling language

AUTISM: Stage 2 Treatments • Weeks 13 to 24 • Depended on early vs. slow response to first

treatment – Percent change from baseline to week 12 based

on 14 communication variables • Early responder: 25% or greater change on

one-half of variables • Slow responder: otherwise

• Early responders continued their intervention from stage 1

• Slow responders received a modified intervention to improve spontaneous communication outcomes

AUTISM: Stage 2 Treatments for Slow Responders

• From Stage 1 JASP + EMT intervention, re-randomized to: – Intensified JASP + EMT (increased dose of

intervention- 3 hours per week) – Augmented JASP + EMT (addition of SGD

with 2 hours per week) • From Stage 1 JASP + EMT + SGD, assigned

to: – Intensified JASP + EMT + SGD (increased

dose of intervention- 3 hours per week)

SMART Results: Primary Aim • Primary Aim: main effect of stage 1 treatment • Starting with JASP + EMT + SGD treatment

initially led to greater total number of spontaneous, communicative utterances (TSCU) at week 24 – Average TSCU for JASP + EMT + SGD

intervention was 61.9 utterances vs. 40.3 utterances for JASP + EMT intervention

– About double the rate of communicative utterance per minute from baseline

SMART Results: Secondary Aim • Secondary Aim: comparison of adaptive

interventions • Greatest (TSCU) at week 24 of three adaptive

intervention was: – JASP + EMT + SGD intensified JASP +

EMT + SGD – Among slow responders

• From other two adaptive interventions: – Augmented JASP + EMT with SGD led to

greater TSCU compared to intensified JASP + EMT • Mean difference in outcome not significant

Some References • Auyeung, et al. “Sequential Multiple Assignment Randomized

Trial Design of Neurobehavioral Treatment for Patients with Metastatic Malignant Melanoma Undergoing High-Dose Interferon-Alpha Therapy.” Clinical Trials: 480-490. 2009.

• Collins, et al. “The Multiphase Optimization Strategy and the Sequential Multiple Assignment Randomized Trial: New Methods for More Potent eHealth Interventions.” Am J Prev Med: S112-S118. May 2007.

• Kasari et al. “Communication Interventions for Minimally Verbal Children with Autism: A Sequential Multiple Assignment Randomized Trial.” Journal of the American Academy of Children and Adolescent Psychiatry, 53:6. 635-646. June 2014.

• Murphy, et al. “Developing Adaptive Treatment Strategies in Substance Abuse Research.” Drug Alcohol Depend: S24-S30. May 2007.