Objective: To investigate the sensitivity and specificity of two retropulsion tests for identifying individuals with Parkinson’s disease that will fall and/or experiencing near falls.

Background: People with PD have an increased risk for falls and experiencing near falls. They are particularly unstable backwards, and different retropulsion tests exist. Item 30 of the Unified Parkinson’s Disease Rating Scale (UPDRS) is most common, involving an expected shoulder pull. Others advocate using an unexpected shoulder pull, e.g. the Nutt Retropulsion test (NRT).

Methods: The study included 104 people with PD (mean age and PD-duration, 68 and 4.7 years, respectively) visiting a neurological clinic during 2006–2011. Those >80 years of age, requiring support in standing or did not understand the instructions were excluded. UPDRS and NRT assessments were conducted in the “on” condition. Participants then registered all falls and near falls by using a diary for six months. Based on this, participants were defined as “stable” (no falls/near falls) or “unstable” (1 fall / near fall). Sensitivity, specificity, positive and negative predictive values were calculated.

Results: Mean (SD) score for UPDRS III was 14.5 (8.1). Fiftyfive (53%) participants were classified as “stable” and 49 (47%) as “unstable”. Both the NRT and item 30 (UPDRS) scores differed significantly (p = 0.003) between the groups. Mean NRT and item 30 scores were 0.18 (0.51) and 0.33 (0.61) for the “stable” group versus 0.63 (0.88) and 0.71 (0.74) for the “unstable” group. Sensitivity/specificity were 37%/87% (NRT), and 55%/75% (UPDRS item 30). Positive/negative predictive values were 72%/61% (NRT) and 66%/65% (UPDRS item 30).

Conclusions: In this mildly affected sample, both NRT and item 30 had low sensitivity in detecting prospective falls and/or near falls over six months. Our findings speak against using either of these tests alone for this purpose and support previous recommendations of using multiple tests when targeting balance problems in people with PD.