Push-and-release test predicts Parkinson fallers and nonfallers better than the pull test: comparison in OFF and ON medication states.

The aim of this study was to compare the Pull Test--Retropulsion Test and the Push and Release Test (P&R) as regards their ability to predict Parkinson (PD) fallers and nonfallers in relation to their medication state. Eighty-two PD patients participated in the study. Fallers (N = 36) and nonfallers (N = 46) were grouped on the basis of their fall history. Fallers were those who had fallen at least once within the last 6 months. The two groups were compared on the basis of the patients' performance in the Pull and the P&R tests, both in their "OFF" and "ON" medication state. The overall accuracy of the classification of PD patients as fallers and nonfallers was determined by means of binomial logistic regression (BLR) and the analysis of the "area under the receiver operating characteristics curve" (AUC). In the OFF medication state, the statistical analysis revealed that the Pull Test was accurate (methods BLR (AUC)) in 85.4% (0.87) of cases and the P&R Test in 86.6% (0.90). In the ON medication state, the Pull Test was only 76.8% (0.78) accurate, while the P&R Test was 87.8% (0.87) accurate. Both clinical tests are valid and relatively equivalent when assessing patients in their OFF medication state; however, the P&R Test is more accurate than the Pull Test in the ON state. This indicates that it is more widely applicable in clinical practice.