A mom sits in your office with her 12-year-old. He stopped going to school in October. Stomachaches every morning, a meltdown at the door, and now he won’t sleep at his grandmother’s house anymore either. She’s done the reading. She’s heard medication is dangerous for kids. She’s also been on a therapist waitlist for four months.
And she asks you the question you’re going to hear a hundred times in practice: “Should we start with therapy or medicine?”
If you learned this material the standard way, your gut says therapy first. CBT for mild to moderate, add an SSRI if it’s severe or therapy isn’t enough. That’s the sequence most of us were taught. A new trial in the September 2026 issue of The American Journal of Psychiatry just put that sequence to a real-world test. The results are going to change how you talk to families, and they carry some quiet lessons for your boards too.
The question nobody had actually tested
Here’s something that surprises a lot of students. For all the confidence we have about pediatric anxiety treatment, the evidence on which treatment to start with has been thin.
When the American Academy of Child and Adolescent Psychiatry (AACAP) released its Clinical Practice Guideline on pediatric anxiety disorders in 2020, it recommended CBT and SSRIs and suggested SNRIs and combination treatment as additional options. But as Dr. A. Reese Abright explained in an interview with The Carlat Child Psychiatry Report, the evidence was too limited for the guideline to recommend sequencing treatment by severity. It simply couldn’t tell you what to do first, or what to do next if the first thing didn’t work.
The big study everyone cites is CAMS, the Child/Adolescent Anxiety Multimodal Study. CAMS found that sertraline plus CBT beat either one alone through 24 weeks. The Carlat Child Psychiatry Report has pointed out the catch, though: that advantage for combination treatment didn’t hold up at longer-term follow-up. And like a lot of classic trials, CAMS enrolled a relatively advantaged group of kids with fewer complicating conditions than you’ll see in a community clinic.
So the practical question stayed open. Therapy or medicine first? And if the first one only half works, do you switch, or do you add?
What the SMART trial did
The study is called “A Pragmatic SMART Study of Medication and CBT Sequencing in Pediatric Anxiety Disorders.” It was led by Dr. Bradley Peterson and Dr. Amy E. West at Children’s Hospital Los Angeles and funded by the Patient-Centered Outcomes Research Institute (PCORI), not a drug company. The American Psychiatric Association highlighted it as one of the featured papers in the September issue of AJP.
SMART stands for Sequential Multiple Assignment Randomized Trial. It’s a design built to answer “what do I do first, and what do I do second?” rather than “does drug X beat placebo?” Here’s how it worked:
- 316 youth, ages 8 to 17, from Los Angeles County, all with anxiety disorders.
- Stage 1 (12 weeks): randomized to fluoxetine or CBT.
- Stage 2 (12 weeks): a second randomization. Kids either continued their initial treatment or had the other treatment added, creating a combination.
- No placebo arm. Every child got an active, evidence-based treatment.
- The primary outcome was the SCARED (Screen for Child Anxiety Related Emotional Disorders), with impairment tracked by the Child Anxiety Impact Scale.
What makes this trial different is who was in it. According to CHLA, participants had severe anxiety, high levels of socioeconomic disadvantage, and co-occurring conditions like depression, school problems, and substance use. Child psychiatrist Dr. Steven Pliszka, reviewing the paper in detail, noted that the sample was heavily Hispanic, with Black youth and other groups well represented, that most participants also had depression, and that roughly half had ADHD. More than 10,000 people were screened to get to those 316.
In other words, these weren’t the “clean” research kids. These were the kids you’re actually going to see.

What they found
The headline: no single approach was clearly superior. Symptom reduction was similar across fluoxetine, CBT, combination treatment, and the different sequences.
There were some modest signals worth knowing:
- Starting with CBT had a slight edge over starting with fluoxetine. The researchers were careful to say this doesn’t make starting with medication a bad choice.
- For kids who started with CBT, adding fluoxetine at week 12 increased the likelihood of improvement. That CBT-then-combination path was the one sequence that pulled ahead on a subset of measures.
- Non-Hispanic white youth did better staying on fluoxetine throughout, while minority youth improved more when shifted to combination treatment.
Dr. Peterson put it plainly in CHLA’s summary of the findings: “It’s perfectly reasonable to begin with medication if that’s what is accessible and is what works for the patient.” In the accompanying AJP editorial, Dr. Melissa Brotman leaned toward therapy first, based on that CBT-then-combination signal. Dr. Pliszka’s response was measured. He agreed with her reasoning but didn’t think the finding was strong enough to tell a family they must start with CBT.
That’s a healthy disagreement between smart people reading the same data. Get comfortable with those. You’ll live in them for your whole career.
The sobering part: these kids were hard to treat
Don’t skim past this. It’s the finding that’ll shape your expectations in practice.
Per Dr. Pliszka’s breakdown, both groups started with a mean youth SCARED score around 42. The clinical cutoff is 25. At week 12, both groups had dropped to about 37. By week 24, the average was about 29, which means many kids were still in the clinical range. Overall, youth-reported SCARED scores fell by 31.7% across the trial. That’s real and statistically significant. But under the study’s strict remission definition, only a small handful of the 316 participants reached full remission by week 24.
Dr. Peterson also noted that symptom improvement kept going for a full year after the six-month trial ended. PCORI has funded six more years of follow-up.
Why does this matter for you? Because you’re going to have families who expect a kid to be “fixed” in six weeks. You’ll need to set expectations honestly: anxiety in kids usually improves slowly, partial response at three months is normal, and it isn’t failure. It’s the point where you decide whether to add something.
What about side effects?
This is the section to bookmark for the next time a parent tells you SSRIs are too dangerous for kids.
Fluoxetine was started at 10 mg daily and adjusted flexibly up to 80 mg. The most common dose ended up being 20 mg. Dr. Pliszka went into the supplementary tables and found something a lot of people wouldn’t expect. On the somatic side effect scale, there was no difference between the fluoxetine group and the CBT group in dry mouth, insomnia, GI complaints, agitation, or weight gain. Anxiety reported as a side effect was actually higher in the CBT group. The one significant difference was decreased appetite with fluoxetine.
Suicidality was monitored with the Columbia-Suicide Severity Rating Scale throughout. Worsening psychiatric events were rare in every arm, and the difference between groups wasn’t statistically significant.
Now, the honest caveat. This trial wasn’t powered to detect rare adverse events, and the FDA boxed warning for antidepressants and suicidal thinking in people under 25 still stands. You’ll still counsel families about it, and you’ll still monitor closely, especially early in treatment and after dose changes. What this data gives you is context. Many of the “side effects” parents worry about show up just as often in kids doing therapy, because they’re symptoms of anxiety itself.

What to lock in for boards
Your exam isn’t going to ask you about a 2026 SMART trial by name. But this study touches several things that absolutely do show up:
- First-line treatment for pediatric anxiety disorders is CBT and/or an SSRI. Benzodiazepines are not first-line in kids. The AHRQ review behind the AACAP guideline found insufficient evidence that they work.
- Know the FDA labeling quirks. Fluoxetine is FDA-approved in youth for depression (age 8 and up) and OCD (age 7 and up), but it isn’t approved for pediatric generalized, social, or separation anxiety. SSRIs are used off-label for those, and that’s evidence-based practice. Duloxetine carries a pediatric GAD indication for ages 7 to 17, and escitalopram gained a pediatric GAD indication in 2023. Questions love to test the gap between “approved” and “evidence-based.”
- Boxed warning: all antidepressants carry a warning for increased suicidal thoughts and behaviors in children, adolescents, and young adults under 25. Expect questions about monitoring frequency and family education.
- Screening tools: the SCARED is a free, widely used pediatric anxiety measure with parent and child versions. The GAD-7 is also used with adolescents. The AACAP guideline encourages brief, free rating scales to track response over time.
- Research design: a SMART design re-randomizes participants at decision points, so it tests treatment strategies rather than single treatments. A pragmatic trial trades tight control for real-world generalizability. That’s external validity, a favorite exam concept.
How this changes the conversation in your office
Let’s go back to that mom and her 12-year-old.
The old script sounded like a rule: therapy first, and we’ll talk about medicine if that doesn’t work. The new script sounds more like a partnership: “Both CBT and medication work for anxiety in kids. Starting with therapy may have a small edge, and if he’s only partly better after about three months, adding the other one is a good next step. But the best first treatment is the one we can actually start. If the therapy waitlist is four months and he’s missing school now, starting fluoxetine is a reasonable, evidence-based choice.”
That’s shared decision-making backed by data. It respects the family’s values, and it acknowledges access, which for a lot of your future patients is the deciding factor. Dr. West said it well: treatment choice can depend on “family preference, accessibility, and other individual or contextual factors.”
It also means you’ll need to be good at both halves. You may not deliver CBT yourself, but you need to know what good CBT looks like, how to find it, and when a referral isn’t working. Dr. West’s team is now studying how well CBT holds up when it’s delivered by clinicians with varied experience in community settings, because CBT is only as good as the person providing it.
The bigger lesson
A lot of psychiatry education teaches you to find the right answer. Real practice asks you to find the right answer for this kid, in this family, with these resources. The SMART trial is a good reminder that sometimes the evidence says, “Several doors lead to the same room.” Your job is to help the family pick the one they can walk through today, and to know which door to open next if the first one sticks.
That’s not a lower standard. It’s a more honest one. And it’s exactly the kind of thinking that separates a prescriber who follows algorithms from one who treats people.
You’re building that judgment right now, one study and one practice question at a time. Keep going.
Want to lock in pediatric psychopharmacology, boxed warnings, and research design before test day? Head over to PMHNPReview.org for board-focused practice questions and review built for future psychiatric nurse practitioners. You’ve got this.
