Generalizability and reproducibility of functional connectivity in autism.
Jace B King, Molly B D Prigge, Carolyn K King and 13 others
PMID 31285817WHAT IT FOUND
No functional connectivity method was reproducible across autism research sites.
Large combined autism-control samples showed widespread decreased connectivity in autism, but site-specific results varied and the replication sample did not show the same findings.
Key findings
01No single method produced wholly reproducible findings across sites and datasets when age, sex, mean head motion, and site were controlled.
02ABIDE I and the combined ABIDE sample showed widespread decreases in resting-state functional connectivity in autism compared with controls, while ABIDE II had few region pairs and the Utah replication sample had none after correction.
03Features most related to different aspects of autism symptoms were distributed across analysis methods without a clear pattern.
STILL TO COME
How it was doneWhat they found
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
Many scans were removed for motion or quality problems: 419 of 1112 in ABIDE I and 405 of 1114 in ABIDE II. Site results varied strongly. After sites with fewer than 10 participants were dropped, 11 ABIDE I sites and 14 ABIDE II sites remained. The brain parcellation was built from adult imaging data, while many participants were children or adolescents. ABIDE scans were often short and lower resolution, and the reference data from Human Connectome Project were not matched for age and sex. Image processing choices can change connectivity results, and the study used one pipeline. The Utah replication cohort was all male.
Declared interests
National Institute of Mental Health and National Institutes of Health funded the work. The supplied text does not list other conflicts.
The easy way to misread this
Do not read decreased connectivity in autism as a clinical test or as a reason to change therapy. The study found poor reproducibility across sites, and no individual feature was shown to be a sensitive and specific biomarker.