Participant characteristics in autism intervention studies.

Broder-Fingert et al. (2019) wrote a short letter to the editor. They did not run a new experiment. Instead they looked at who gets into autism intervention studies.

They warned that age, IQ, language level, and race often differ from the real-world autism population. These gaps can tilt the results you read.

The letter says many trials leave out kids with low IQ, limited speech, or other diagnoses. When these kids are missing, the study may look more successful than it really is.

In short, the "average" finding may only apply to higher-functioning children.

Trembath et al. (2019) backs this up. Their review of 41 parent-training studies lists 45 child or parent factors that change outcomes. Language level and parent adherence sit at the top of that list.

Tromans et al. (2018) adds size to the worry. They scanned 529 autism RCTs and found most have only about 36 participants. Small, skewed samples plus poor description equals shaky evidence.

Delamater et al. (1986) sounded the same alarm 33 years earlier. They audited older papers and found spotty subject description. The problem is old, but it is still here.

Before you trust a shiny new intervention, flip to Table 1. Check who was actually studied. If the sample is mostly verbal, high-IQ, white boys, think twice before you try the protocol with a non-verbal, lower-IQ, bilingual girl. Ask the question, "Do my clients look like this sample?" If the answer is no, demand more data or pilot carefully with your own measures.