Screening performances of an 8-item UPSIT Italian version in the diagnosis of Parkinson's disease

Hyposmia is a common finding in Parkinson's disease (PD) and is usually tested through the University of Pennsylvania Smell Identification Test (UPSIT). The aim of our study is to provide a briefer version of the Italian-adapted UPSIT test, able to discriminate between PD patients and healthy subjects (HS). By means of several univariate and multivariate (machine-learning-based) statistical approaches, we selected 8 items by which we trained a partial-least-square discriminant analysis (PLS-DA) and a decision tree (DT) model: class predictions of both models performed better with the 8-item version when compared to the 40-item version. An area under the receiver operating characteristic (AUC-ROC) curve built with the selected 8 odors showed the best performance (sensitivity 86.8%, specificity 82%) in predicting the PD condition at a cut-off point of <= 6. These performances were higher than those previously calculated for the 40-item UPSIT test (sensitivity 82% and specificity 88.2 % with a cut-off point of <= 21). Qualitatively, our selection contains one odor (i.e., apple) which is Italian-specific, supporting the need for cultural adaptation of smell testing; on the other hand, some of the selected best discriminating odors are in common with existing brief smell test versions validated on PD patients of other cultures, supporting the view that disease-specific odor patterns may exist and deserve a further evaluation.

Automated summary

This study aims to provide a briefer version of the Italian-adapted UPSIT test that can discriminate between Parkinson's disease patients and healthy subjects. By using statistical approaches and machine learning, 8 items were selected and tested, showing better performance compared to the 40-item version. The selected 8 odors demonstrated higher sensitivity and specificity in predicting the PD condition.