A sweeping population-based cohort study from Sweden's Karolinska Institutet, published in The BMJ (DOI: 10.1136/bmj-2025-084164), presents compelling evidence that the long-accepted male predominance in autism spectrum disorder (ASD) may be, in substantial part, an artifact of diagnostic bias rather than a reflection of true sex-based prevalence differences.
The study is notable for both its scale and duration. Researchers analyzed national health records covering approximately 2.7 million individuals born between 1985 and 2022, with follow-up periods extending up to 37 years. This prospectively collected birth cohort design represents one of the most comprehensive epidemiological investigations of autism prevalence to date.
The findings directly challenge the widely cited 4:1 male-to-female ratio in autism prevalence. The data reveal a complex, age-dependent pattern: while males are diagnosed at substantially higher rates during childhood, females experience a steady, persistent increase in diagnosis rates throughout adolescence. By approximately age 20, the male-to-female incidence ratio converges toward 1:1. This convergence pattern held across successive birth cohorts, though the overall ratio has been decreasing over calendar time as well.
In absolute terms, the study identified 78,522 individuals (2.8% of the total cohort) who received an autism diagnosis during the follow-up period. The mean age at diagnosis was 14.3 years, though this figure masks significant gender differences. Male diagnosis rates peaked between ages 10-14 at 645.5 per 100,000 person-years, while female rates peaked later, between ages 15-19, at 602.6 per 100,000 person-years. The temporal lag between male and female peak diagnosis rates — approximately 5 years — is itself a key finding, suggesting systematic delay in female identification rather than lower true prevalence.
The decreasing male-to-female ratio across successive birth cohorts is particularly instructive. Earlier cohorts (those born in the 1980s and early 1990s) exhibited substantially larger gender disparities in diagnosis. More recent cohorts show progressively smaller gaps, consistent with improved clinical awareness of female autism presentations over time. The authors concluded that "the male to female ratio for autism has decreased over time and with increasing age at diagnosis" and may "no longer be distinguishable by adulthood."
Several methodological and clinical factors have historically contributed to the underdiagnosis of autism in females. The diagnostic criteria codified in the DSM and ICD were developed and validated predominantly on male samples, creating an inherent male-normed framework. The "female autism phenotype" — now increasingly recognized in clinical literature — includes characteristics that may fall below traditional diagnostic thresholds: superior social camouflaging and masking abilities, more internalized symptomatology (anxiety, depression) versus the externalized behaviors (aggression, hyperactivity) more common in males, fewer stereotyped repetitive behaviors, more socially normative special interests, and more typical language development trajectories. These presentation differences mean that existing screening instruments, many of which emphasize male-typical features, may lack sensitivity for female presentations.
The compensatory strategies that many autistic girls develop — collectively termed "masking" or "camouflaging" — are thought to delay diagnosis by obscuring core social communication difficulties. Research suggests that masking carries significant psychological costs, including chronic exhaustion, identity confusion, and elevated rates of anxiety and depression. The adolescent peak in female diagnoses aligns with a developmental period when social demands intensify, peer relationships become more complex, and the cognitive resources required for sustained masking may exceed capacity.
The study's limitations warrant careful consideration. As an observational study utilizing administrative health records, it can identify associations but cannot establish causal mechanisms for the observed convergence in gender ratios. Several potential confounders remain uncontrolled: comorbid neurodevelopmental conditions (particularly ADHD, which has its own gender diagnostic disparities), intellectual disability (which affects recognition patterns), genetic loading, environmental exposures, and family psychiatric history. The study cannot definitively distinguish between three possible explanations for the narrowing gender gap: (1) true increases in female autism prevalence, (2) improved diagnostic recognition of existing female autism, or (3) changes in diagnostic criteria that capture a broader phenotypic spectrum.
Furthermore, the reliance on formal clinical diagnosis as the outcome measure introduces ascertainment bias — individuals who never present to clinical services, particularly those with subtler presentations, are systematically excluded. This is likely to disproportionately affect females, suggesting the true prevalence gap may be even smaller than observed.
Despite these methodological caveats, the clinical and public health implications are substantial. If autism is indeed distributed more equally across genders than prevailing estimates suggest, current diagnostic and screening practices may be systematically failing to identify millions of girls and women. The consequences of missed or delayed diagnosis include absence of early intervention, inappropriate treatment for misdiagnosed conditions, social marginalization, and the cumulative psychological toll of unrecognized neurodevelopmental difference. These findings strengthen the case for sex-sensitive diagnostic tools, clinician training on female autism presentations, and screening protocols that account for the masking phenomenon.