Age, sex or medication state differ between groups
Symptom. Patients older and sex-skewed relative to controls; patients recorded on medication.
Symptom
A “biomarker” that separates patients from controls also separates old from young, men from women, or medicated from unmedicated in the same data — and the groups differ on those variables. The effect is in measures known to vary with age (alpha frequency, aperiodic slope, absolute power), with sex (skull thickness and hence amplitude, head size), or with the drug (dopaminergic medication attenuates beta signatures in Parkinson’s disease). Matching the groups on the confound, or adjusting for it, shrinks the effect; a replication in a better-matched cohort finds less.
Cause
Clinical groups are rarely sampled the way controls are. Patients are older on average, drawn from a clinic rather than a community, more often of one sex for many conditions, and recorded while treated. Each of those variables has its own, well-documented effect on the EEG, and when it is unbalanced between groups the group comparison measures it too. The effect is compounded when the variable also affects data quality (older participants move more; some medications alter arousal), so that artifact rates differ between groups and any residual artifact becomes part of the “effect”. Medication is the most specific case: recording patients on treatment measures the treated brain, and the comparison with untreated controls confounds disease with drug.
Detect
- Tabulate age, sex, medication state, handedness and any other covariate per group, with means, spreads and counts; report the imbalance rather than assuming matching.
- Test whether the candidate biomarker correlates with the covariates within the control group; a measure that tracks age in controls will track an age imbalance between groups.
- Compare artifact rates and data quality between groups before comparing the measure.
- Read the dataset description for medication state and recruitment source; the catalog records them where the descriptor does.
Fix
- Design: age- and sex-matched recruitment; record patients off medication where clinically possible, or record both states; document the covariates in
participants.tsv. - Analysis: model the covariates (age, sex, medication) rather than only matching on them, with enough data to estimate their effects (L6.2); report the group effect with and without adjustment.
- Interpretation: state the confound as a limitation in the same sentence as the effect; a biomarker that cannot be separated from age is an age marker until shown otherwise.
- Replication: seek a cohort in which the confound is absent or reversed before generalizing.
Example
Two datasets in the site’s directory carry this confound as documented facts (from §9.2 and §10.9 of the spec, copied from the catalog):
| Dataset | Documented imbalance | Consequence |
|---|---|---|
ds-iowapd | 100 PD patients versus 49 controls; sex 68/32 versus 26/23 male/female; patients recorded on dopaminergic medication, which attenuates beta signatures | group imbalance, sex imbalance, and a disease-versus-drug confound in any resting biomarker (L6.3, L7.6) |
ds-aszed | Patients older and sex-skewed relative to controls, per its pilot subset (the full release is not itemized) | age and sex confounds in schizophrenia-versus-control contrasts (L7.6) |
No figure is generated for this entry: the confound lives in the participant table, not in a trace. The Level 6 lesson on mixed models shows how to include such covariates, and L7.6 returns to ds-iowapd with the numbers above.