What this is: This reproduces case-control cortical thickness effect sizes from the ENIGMA-ADHD Working Group (Hoogman et al., 2019, American Journal of Psychiatry), using publicly available summary statistics.
What's original: the adult-vs-adolescent age-group comparison and the interactive visualizations are my own extensions of the published analysis; I also re-applied the FDR correction myself.
Tools: Python (pandas, scipy/statsmodels for FDR correction, matplotlib, Plotly)
Hover over bars for exact values. Open in a new tab →
| Region | Cohen's d | p | FDR q |
|---|---|---|---|
| lateraloccipital | 0.140 | 0.020 * | 0.437 |
| caudalanteriorcingulate | -0.110 | 0.060 | 0.437 |
| cuneus | 0.110 | 0.060 | 0.437 |
| lingual | 0.110 | 0.070 | 0.437 |
| posteriorcingulate | -0.110 | 0.050 | 0.437 |
* p < 0.05 uncorrected ** survives FDR q < 0.05
Each bar shows Cohen's d for cortical thickness differences between ADHD patients and controls in adults. Positive values mean thicker cortex in ADHD; negative means thinner.
Data source: ENIGMA Toolbox (MICA-MNI). Analysis code and full pipeline: this repository.
The chart above uses adult data only. But ENIGMA also has a separate adolescent cohort - so I compared the two to see whether ADHD affects the same brain regions regardless of age, or whether the pattern shifts.
Hover over points for exact values. Open in a new tab →
| Region | Adult d | Adolescent d | Diff |
|---|---|---|---|
| parahippocampal | 0.100 | -0.050 | +0.150 |
| bankssts | 0.010 | 0.160 | -0.150 |
| lateralorbitofrontal | 0.030 | -0.120 | +0.150 |
| lingual | 0.110 | -0.030 | +0.140 |
| cuneus | 0.110 | -0.010 | +0.120 |
Top 5 regions by largest adult–adolescent divergence. ENIGMA-ADHD reports bilateral values, so left and right hemispheres are identical in the source data.
Each point is one brain region. If a region were affected identically in both age groups, it would sit on the red dashed line. Across all 34 regions, the correlation between adult and adolescent effect sizes is only weak (r = 0.38). Because almost no regional effects reach significance in either group, much of this divergence may reflect sampling noise rather than true age-dependent change.
The clearest example: bankssts shows almost no effect in adults (d ≈ 0.01) but a larger effect in adolescents (d = 0.16). The opposite pattern shows up in the parahippocampal region. These are single effect sizes from moderate samples, so they should be treated as hypotheses, not findings.
Caveat: this compares two different groups of people (cross-sectional), not the same patients tracked over time - so it shows that adult and adolescent ADHD populations differ structurally, not that any individual's brain changes this way as they age. Differences in scanner sites and sample sizes between cohorts could also contribute.
ENIGMA-ADHD reports bilateral thickness values for 34 cortical regions, so correction was applied across 34 tests. Only 1 of 34 regions is nominally significant (p < 0.05) in adults, and after re-applying a Benjamini-Hochberg FDR correction, none survive at q < 0.05. Same in adolescents: 2 regions nominal, 0 surviving FDR. This mirrors the original ENIGMA-ADHD finding: individual cortical effects are small and diffuse, which is precisely why consortium-scale sample sizes (thousands of scans) are needed to detect them reliably.
Analysis code: analyze.py