Clinicians treating an anxious child face two questions in sequence: what to start with, and what to do when it does not work. This pragmatic sequential multiple assignment randomised trial addressed both. Three hundred and sixteen young people aged 8 to 17 with a DSM-5 anxiety disorder, recruited from primary care and mental health clinics, were randomised first to fluoxetine or exposure-based CBT for 12 weeks, then — for those not in remission — randomised again to continuing the initial treatment optimised, or adding the other modality. The cohort had severe anxiety, high sociodemographic disadvantage and frequent co-occurring diagnoses. The primary outcome was the youth-reported SCARED score at 24 weeks.
Everything worked, and nothing worked better. Youth-reported scores fell 31.7% overall. Initial treatment made no significant difference, with CBT holding a non-significant edge (24-week difference in mean change 1.45, 95% CI -2.25 to 5.16). Among week-12 non-remitters, combination therapy was no better than continuing monotherapy (difference -2.74, 95% CI -6.53 to 1.05). One sequence — CBT first, then combination — separated from the others on a subset of measures.
The authors also report that non-Hispanic White young people did better starting and staying on fluoxetine, while racial and ethnic minority young people did better transitioning to combination therapy after week 12. That is a subgroup finding in a trial of 316 patients and should not drive decisions; it is worth noting as a question rather than an answer.
The useful conclusion is permissive rather than prescriptive. When several approaches produce similar results, the right one is the one the family will actually engage with. A child whose parents are wary of medication should start with CBT; a family who cannot get to weekly therapy sessions should not be denied treatment for that reason. In Indian practice, where access to trained paediatric CBT therapists is very limited outside major cities, this is genuinely reassuring: starting with fluoxetine is not a second-best compromise, and it does not condemn the child to a worse outcome.
What the trial does not support is the reflex of adding the second modality when the first has not worked by three months. That was tested and it did not help.
- Start with whichever of fluoxetine or CBT the family will engage with; outcomes were similar.
- Where CBT is unavailable, starting with fluoxetine is not a compromise on outcome.
- Do not reflexively add the second modality at 12 weeks in a non-remitter — that was tested and did not help.
- Optimising the initial treatment was as good as combining; revisit dose, adherence and exposure quality first.
- Treat the ethnicity subgroup findings as hypothesis-generating, not as a basis for treatment selection.
The statistics, in plain English
A SMART design randomises twice — once at the start and again at the point where treatment is judged insufficient — which is what allows a sequence to be tested rather than just a treatment. The results are null with reasonably tight intervals: the initial-treatment difference of 1.45 with an interval of -2.25 to 5.16 sits comfortably around zero on a scale where the whole cohort improved by about a third. Note that all 316 patients were split across two randomisations, so the second-stage comparisons rest on far fewer people than the first, and the finding that one sequence separated 'on a subset of measures' is the kind of result that appears when many outcomes are examined.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for psychiatry, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free