Olfactory Testing in Parkinson Disease and REM Behavior Disorder: A Machine Learning Approach

Area of Science:

• Neuroscience
• Clinical Neurology
• Diagnostic Tools

Background:

• Olfactory dysfunction is an early non-motor symptom in Parkinson disease (PD) and isolated REM sleep behavior disorder (iRBD).
• Accurate and efficient screening tools are needed for clinical environments to identify olfactory deficits in these populations.

Purpose of the Study:

• To develop and validate an abbreviated olfactory test for use in busy clinical settings.
• To assess the utility of this test in stratifying individuals with prodromal (iRBD) and manifest Parkinson disease (PD).

Main Methods:

• A 3-item Sniffin' Stick odor identification model was developed using random forests on a large PD cohort (n=890) and controls (n=313).
• The model's predictor importance ranked the top 3 sticks (anise/licorice/banana) for distinguishing poor vs. good smell ability.
• Validation was performed in independent PD (n=452) and iRBD (n=241 and n=37) datasets.

Main Results:

• The 3-stick model achieved an AUC of 0.95 in the development dataset, with high sensitivity (90%) and specificity (92%) for detecting impaired olfaction.
• Internal and external validation confirmed robust performance (AUCs ≥0.90).
• Combined with an RBD screening questionnaire, the 3-stick model differentiated iRBD from controls with 65% sensitivity and 100% specificity.

Conclusions:

• A brief 3-Sniffin'-Stick model demonstrates significant potential as a screening tool for olfactory dysfunction in PD and iRBD.
• This abbreviated test can aid in the clinical stratification of patients based on olfactory deficits.
• Class III evidence supports its utility in identifying individuals with iRBD.