Conducting genetic epidemiology studies of autism spectrum disorders: issues in matching.

Szatmari et al. (2004) wrote a guide for researchers who run family-genetic studies of autism. They explain how to pick control families so the results truly reflect autism genes, not other differences.

The paper lists common traps. For example, if autism families differ in age, IQ, or income, those factors can look like genetic signals.

The authors did not run a new experiment. Instead they show that sloppy matching can drown real gene effects in noise.

They give checklists for age, sex, IQ, and social class so future studies can avoid false leads.

Three other 2004 papers echo the same worry. Jarrold et al. (2004) tell us to match on control-task scores, not just IQ. Ganz et al. (2004) warn against age-equivalent scores and loose p-value rules. Together they form a wake-up call from that year: stop declaring groups “matched” after only one test.

Later work extends the message. Facon et al. (2011) say we must also check variance and shape of scores, not just the mean. Flapper et al. (2013) add effect-size and variance-ratio rules for wider developmental disabilities. These updates do not contradict Peter et al.; they simply add new screens to the same sieve.

Lancioni et al. (2011) shift the lens from genes to biomarkers yet keep the core plea: prove your groups are truly alike before you claim a marker is valid.

If you consult on grant design or peer-review journal submissions, use this paper set as a quick checklist. Ask authors how they matched participants and whether they checked variance, not just averages. A short table showing age, IQ, and SES effect sizes with variance ratios can save a study from a fatal Reviewer 2 comment.