Newly diagnosed autism and ADHD carry less genetic risk each year - which is an argument about criteria

Design

Population-based case-cohort study with polygenic score regression against simulated scenarios, Denmark

Population

17,071 individuals with incident autism diagnoses and 20,111 with ADHD, diagnosed 1994-2016

Primary outcome

Change in mean polygenic scores by year of incident diagnosis

Effect

ADHD score -0.06 SD per decade (95 per cent CI -0.09 to -0.03) in ADHD cases; autism score -0.07 SD per decade (-0.10 to -0.04) in autism cases

Rates of autism and ADHD diagnosis have climbed for three decades, and the usual explanations are awareness, reduced stigma and changing practice. This Danish population-based case-cohort tested that directly by asking whether the people receiving those diagnoses have changed genetically.

Among 17,071 individuals with incident autism diagnoses and 20,111 with ADHD, made between 1994 and 2016, polygenic risk fell steadily with year of diagnosis. For ADHD cases the ADHD polygenic score dropped by 0.06 standard deviations per decade (95 per cent CI -0.09 to -0.03, P = 0.001), and scores for autism, bipolar disorder and schizophrenia fell too. For autism cases, the autism score fell by 0.07 standard deviations per decade (-0.10 to -0.04, P < 0.001), with similar declines for bipolar disorder, schizophrenia and educational attainment. Comparing these trends against simulations of different possible drivers, the pattern fitted broadening diagnostic criteria rather than the emergence of new risk factors.

This is the most direct evidence yet against the proposition that something in the environment is generating more autism and ADHD. It is also a finding about populations, not patients: a lower average polygenic score in a cohort says nothing about any individual, and a diagnosis made under broader criteria is not a false one - the impairment that brought the person to services is real, whatever the genetic loading. What it does mean is that the average newly diagnosed patient today is less severely loaded than the average patient diagnosed in 1994, which is worth remembering when older prognostic literature is applied to a current clinic.

• Do not treat rising diagnosis rates as evidence of a new environmental cause; the genetic data do not support it.
• Apply older prognostic and outcome literature cautiously - it describes a more heavily loaded cohort.
• Never infer anything about an individual patient from a polygenic score trend across a population.
• A diagnosis made under broadened criteria still describes real impairment requiring real support.
• Note this is a single national dataset with Danish ancestry adjustment; transfer to other populations is untested.

The statistics, in plain English

A decline of 0.06 standard deviations per decade is a small shift in a population mean, detectable only because the cohort is very large; it says nothing about where any individual sits. Polygenic scores capture a fraction of heritability, so a falling score means the diagnosed group is less genetically enriched, not that genetics matters less. Comparing observed trends against simulated scenarios is a strong design here, because it tests competing explanations against each other rather than merely rejecting a null.

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.