Sometimes, a finding does not overturn a field but quietly reframes it. That is how I read the recent population-based study in The BMJ by Caroline Fyfe and colleagues on time trends in the male-to-female ratio in autism diagnoses.

For decades, the story seemed relatively stable. Autism spectrum disorder was said to be three to four times more common in boys than in girls. That ratio became part of the common knowledge around autism. It fuelled biological theories, including the female protective effect, and shaped how clinicians, teachers and researchers thought about identification.

This new Swedish cohort study, covering nearly 2.8 million children born between 1985 and 2020, complicates that picture in an important way. The authors disentangle age at diagnosis, calendar period and birth cohort. That matters because if diagnoses rise over time, and if girls tend to receive a diagnosis later than boys, a simple overall ratio can be misleading.

The overall increase in autism diagnoses is striking. Among children aged 10 to 14, incidence rose roughly tenfold between 2000 and 2022. That finding aligns with what we have seen internationally. Broader diagnostic criteria, growing awareness and better access to services all play a role. The study does not claim a single explanation, and rightly so. Autism is defined behaviourally, and when the criteria shift, numbers move.

The more intriguing finding concerns the male-to-female ratio. In early childhood, the ratio remains clearly skewed towards boys. Around age 10, the cumulative male-to-female ratio is still about three to one. But from adolescence onwards, the picture changes. For more recent diagnostic periods, the ratio steadily declines with age. By age 20, for diagnoses recorded in 2022, the cumulative ratio approaches 1.2 to 1. Under projected trends, it may reach parity.

In plain terms, girls appear to catch up. Not because boys are suddenly less likely to receive a diagnosis, but because increasing numbers of girls and young women are being identified later in adolescence and early adulthood.

That finding does not immediately tell us why this happens. The authors are careful here, and that is a strength. Several explanations remain plausible. Diagnostic criteria have broadened. Awareness of how autism presents in girls has improved. Camouflaging behaviours may delay identification in childhood. Co-occurring psychiatric diagnoses may initially obscure autistic traits. It is also possible that referral patterns differ by sex, particularly in adolescence, when young women may seek mental health care more frequently.

One should also keep a few caveats in mind. Register-based data are powerful but blunt instruments. They tell us who received a diagnosis, not how traits manifested or how clinical judgement evolved in individual cases. Outpatient data were only fully included from 2001 onwards, which may have influenced age at diagnosis in earlier cohorts. And projections beyond the observed data, even when statistically well modelled, remain projections.

Still, the core message stands. The historically cited male-to-female ratio of four to one may be less a biological constant and more a moving target shaped by diagnostic practice, awareness, and life-course effects.

For education and youth services, that is not a trivial shift. If girls are identified later, often after years of internalising difficulties or misattributed symptoms, then support systems need to adapt. Early screening, teacher training and mental health services should be alert to different presentations, not only the more stereotypical male profile.

The study does not dismiss biological explanations. Nor does it claim that sex differences disappear. It shows something more subtle and, in my view, more interesting. The ratio changes with age and over time. That suggests that at least part of the historical gap reflects when and how we recognise autism, not only its distribution in the population.

In research, we often search for stable truths. This study reminds us that some truths are time-bound. When diagnostic systems evolve, our numbers evolve with them. The task is not to defend old ratios, but to understand what the changing ones reveal about our practice.