Challenges in evaluating psychosocial interventions for Autistic Spectrum Disorders.

Lord et al. (2005) looked at every psychosocial autism study they could find. They asked one simple question: can we trust the results?

The team read hundreds of papers and wrote a consensus report. They checked how many used random assignment, blind raters, or the same outcome measure.

Most studies were small, short, and used different yardsticks. Few used random assignment or blind raters.

The authors said the field needs more RCTs and one shared yardstick before we can say what really works.

Fernell et al. (2014) took the same plea and aimed it at early-screening studies. They agreed we need rigor, but said shorter trials can still count.

Jonsson et al. (2016) later showed that even when RCTs exist, they rarely tell us if the program will work in real schools. They built a checklist so you can spot missing details.

Stewart et al. (2018) pooled 19 parent-training RCTs and found small but real gains. Their meta-analysis is exactly the kind of evidence Lord et al. (2005) said was missing.

If you pick an autism program today, you still face a patchwork of weak studies. Use the 2016 external-validity checklist before you adopt any new package. Push vendors for RCT data and demand they use the same outcome measure your clinic already tracks.