Falling polygenic scores argue that broader criteria, not new risk, drive rising ADHD and autism rates

Design

population-based case-cohort study with polygenic score analysis and simulation comparison

Population

17,071 people with incident autism spectrum disorder and 20,111 with ADHD diagnosed in Denmark, 1994 to 2016

Primary outcome

change in mean polygenic scores by year of incident diagnosis

Effect

ADHD score −0.06 SD per decade (95% CI −0.09 to −0.03); autism score −0.07 SD per decade (95% CI −0.10 to −0.04)

A Danish population-based case-cohort used the iPSYCH2015 resource to look at 17,071 people with an incident autism spectrum disorder diagnosis and 20,111 with attention-deficit/hyperactivity disorder, diagnosed between 1994 and 2016. The question was whether the genetic risk profile of the people receiving those diagnoses changed as the diagnoses became more common, with polygenic scores for psychiatric and cognitive-behavioural traits as the measure, adjusted for age, sex and ancestry.

Scores fell as the years passed. For ADHD, mean polygenic score for ADHD dropped by 0.06 standard deviations per decade of diagnosis year (95% CI −0.09 to −0.03; P=0.001), with parallel falls for autism, bipolar disorder and schizophrenia scores. For autism, the autism score fell 0.07 standard deviations per decade (95% CI −0.10 to −0.04; P<0.001), with similar falls for bipolar disorder, schizophrenia and educational attainment. The authors then compared these trends against simulations of what different explanations for rising incidence would predict.

The observed pattern fits broadening diagnostic criteria: each successive cohort included people with less genetic loading, which is what happens when a threshold moves outward rather than when a new environmental cause appears. A genuine new risk factor would have pushed scores in the other direction or left them flat.

This is a useful thing to be able to say to a worried parent. It is also a caution about reading old research onto new patients: a trial recruited in 1998 studied a more genetically loaded group than the clinic population of 2026, which may be part of why effect sizes look smaller now.

• Use this when families ask what is causing the rise — the evidence points to who gets diagnosed, not to a new exposure
• Expect published effect sizes from older trials to overstate what current, milder cohorts will show
• Do not read a lower polygenic score as a milder illness in an individual — these are group means
• Keep the functional impairment threshold explicit in your own diagnoses
• Note that scores for bipolar disorder and schizophrenia fell too, so this is not specific to the index diagnosis

The statistics, in plain English

A shift of 0.06 to 0.07 standard deviations per decade is small at the level of any individual and would be invisible in a clinic; it is detectable only across tens of thousands of people. The strength of the finding is not its size but its direction and its consistency across several unrelated polygenic scores, which is hard to produce by chance or by a single confounder. The simulation comparison is what carries the causal argument, and it is a modelling argument rather than an observation.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.