The age at which autism is diagnosed is linked to genetic differences that are not all explained by biological sex, according to a new study in Behavior Genetics. The researchers found that the genetic profiles of autism in boys and girls are more similar to each other than those of people diagnosed early and later. For early diagnosis, some associations with educational and cognitive traits remained when sex was taken into account. For late diagnosis, however, it was not possible to separate the two factors.
Autism is a neurodevelopmental disorder characterized by difficulties with communication and social interaction, as well as repetitive behaviors or highly restricted interests. Historically, it has been diagnosed about four times more often in boys, who usually receive a diagnosis earlier than girls. A 2025 study had identified different genetic profiles for early and late diagnosis. The new research examined whether this difference reflects the unequal representation of the two sexes in the respective groups.
The team used a statistical method called genomic structural equation modeling, which examines shared genetic effects across traits. It analyzed summary results from genome-wide association studies for four categories: early diagnosis, late diagnosis, male sex and female sex. Depending on the individual sample, early diagnosis meant diagnosis before an age cutoff of roughly six to nine years, while late diagnosis meant age ten or older. The effective sample size was 33,608 for early diagnosis and 21,775 for late diagnosis, while the data by sex included 15,025 males and 4,845 females with autism, with more than 19,000 controls in each comparison.
The researchers compared the four categories with 68 traits, ranging from psychiatric disorders and cognitive abilities to loneliness and satisfaction with family relationships. As lead author Sophie Breunig, a doctoral candidate at the University of Colorado Boulder, explained to PsyPost, genetic correlations describe the overlap in genetic effects across large population groups. They range from −1 to 1 and cannot be used to diagnose or make predictions about a particular individual. Their value lies in understanding the variation among people who share the same diagnosis.
The genetic correlation between autism in the two sexes was 0.81, indicating strong similarity. Only two of the 68 traits showed clearly different correlations by sex: anorexia nervosa had greater genetic overlap with autism in females, while the tendency to exceed the speed limit when driving showed a negative correlation only in this group. Both correlations remained when early diagnosis was taken into account. The authors also clarify that the similarity of common genetic variants is a different research question from whether the heritability of autism differs between the sexes.
By contrast, the genetic correlation between early and late diagnosis was 0.42, with 31 traits showing different correlations. Late diagnosis had greater genetic overlap with major depression, post-traumatic stress disorder and attention-deficit/hyperactivity disorder, as well as with loneliness and lower satisfaction with personal relationships. Early diagnosis was more strongly associated with higher educational attainment, but also with lower scores on tests of memory and processing speed. Both groups showed a similar positive genetic correlation with general intelligence.
When sex was taken into account, the distinctive associations with early diagnosis remained, and some became stronger. This was particularly true of performance on a symbol-matching test and the portion of educational attainment that was not explained by cognitive ability. The equivalent analysis was not possible for late diagnosis: its genetic signal had a correlation of 0.99 with that of autism in males and 0.91 with that of autism in females. This very large overlap prevented the statistical separation of age at diagnosis and sex.
In an additional analysis, the team combined depression, anxiety and post-traumatic stress into a general emotional distress factor. When it took this factor into account, the genetic correlation between late autism diagnosis and adult ADHD fell from 0.42 to almost zero. The correlation between autism and ADHD also fell more in females, from 0.39 to 0.07, than in males, from 0.51 to 0.23. According to the researchers, this suggests that part of the overlap may relate to broader emotional distress.
The data came exclusively from people of European genetic ancestry, and the sex categories referred to sex recorded at birth. The authors consider the conclusions about distinctive psychiatric associations with late diagnosis tentative, as they could not isolate the effect of sex. They also note that other conditions may mask or resemble autism during diagnosis in adulthood. Larger samples could clarify these relationships and potentially help professionals recognize different support needs.
Comments