Hypothesis
A short subset of odours that Parkinson's patients selectively fail to identify would discriminate patients from controls better than the full 40-item UPSIT or a random set of items.
What failed
Several 12-item subsets beat 12 random items in a discovery cohort of 314 patients and 314 controls, but in two independent validation cohorts (306 patients, 343 controls) they matched the full test without outperforming 12 random items.
Broken assumption
Parkinson's disease impairs identification of specific odours more than others, in a way that is stable across patient groups.
Why
The ranking of which odours discriminate best did not hold across cohorts, which the authors read as evidence against odour-selective smell loss in Parkinson's disease.
From the published paper: Morley et al. npj Parkinson s Disease, 2018. ↗CC BY