New gene targets for OCD and tics; autism diagnoses rising fastest in girls and women; who was that AI tested on? — Morning Medical Update

New gene targets for OCD and tics

Thirty-six large-effect risk genes turned up in an analysis of nearly 4,000 patients.

Researchers have identified 36 genes that substantially raise the risk of obsessive-compulsive disorder and chronic tic disorders, according to a whole-exome sequencing study of 3,964 patients published in Nature Neuroscience. Only four such genes had been established before. The Rutgers-led team analyzed 2,418 parent-child trios alongside 1,734 control trios, roughly doubling the size of prior case samples.

The genes carry large effects, with an average odds ratio of 57, but rare damaging mutations were found in only about 3% to 8% of affected patients. Several overlap with genes already tied to autism, schizophrenia and developmental delay. Jay Tischfield, an emeritus distinguished professor of genetics at Rutgers and a senior coauthor, said in a statement that the genes act in networks rather than individually, which he said makes new therapies easier to design.

Autism diagnoses rising fastest in girls and women

Female incidence climbed 19% a year after 2020 while male rates held steady or fell.

New autism spectrum disorder diagnoses rose sharply among female patients after 2020 while male rates stayed flat or declined, according to a retrospective cohort study of 171,134 patients in a multistate Midwestern health system published in JAMA Network Open. Female incidence climbed 19.0% a year from 2020 to 2024, while male incidence fell 3.8% annually across the 2016 to 2024 period.

The pattern held in children, adolescents and adults, and females were diagnosed a mean 3.4 years later than males, at 15.7 years versus 12.3. Erick Messias, M.D., M.P.H., Ph.D., of Saint Louis University and colleagues wrote that the increase reflects better identification of historically underrecognized presentations rather than a uniform population-level rise. The analysis relied on billing codes and required a clinical visit every year, which skews the cohort toward consistently insured patients.

Who was that AI tested on?

Only 14% of validation studies for commercial radiology AI reported results by sex, age or race.

Most published validation studies of commercially available radiology artificial intelligence products do not report how the software performed across patient subgroups, according to a scoping review of 545 studies covering 252 products published in European Radiology. Researchers at Osaka Metropolitan University found that 392 studies reported subgroup demographic data, but only 77, or 14%, also reported results broken out by sex, age or race and ethnicity.

Reporting has not become more common over time, and the authors found no association with company sponsorship. Applying a statistical power calculation to tuberculosis detection studies, they estimated that 14 of 21 datasets were too small to support a sex-based subgroup analysis. Shannon L. Walston, who led the review, said the commercial validation record is too fragmented to estimate subgroup bias. The authors disclosed a relationship with Medical AI Promotion Institute, Inc.