The use of a color coded probability scale to interpret smell tests in suspected parkinsonism

Smell identification tests (SITs) have been suggested as useful in the differential diagnosis of Parkinson's disease (PD). We have applied the 40 item University of Pennsylvania SIT (UPSIT-40) and/or the 16 item SIT from Sniffin Sticks (SS-16) to 193 nondemented PD patients and 157 controls and used logistic regression analysis to associate the SIT result with the probability of an individual patient having PD ("PD probability"). Reliability measures (95% CI) using the clinical diagnosis as a gold standard and a dichotomized result of the smell test into high (50% or more) or low (<50%) "PD probability" were: sensitivity 85.0% (78.8-89.7%), specificity 84.6% (77.3-89.9%) for the UPSIT-40; sensitivity 90.4% (83.5-94.7%), specificity 85.5% (76.2-91.7%) for the SS-16. Based on these findings we have created color coded visual tools (PD probability rulers) and applied them to 39 clinically uncertain parkinsonian syndromes (CUPS) patients who had been investigated with dopamine transporter SPECT scanning using [123-I]-FP-CIT SPECT (DaTSCAN) for suspected Parkinson's disease. In 32 of 36 CUPS cases (88.9%, kappa = 0.72) the probability ruler predicted the result of the DaTSCAN. We suggest smell tests could be used routinely in challenging cases where there is diagnostic uncertainty and help inform decision making relating to the need for neuro imaging.