In this study, we investigated the effect of dopaminergic medication on reactive stepping responses to forward and backward balance perturbations in patients with moderately severe Parkinson's disease (PD). Twelve PD patients, Hoehn and Yahr stage ranging from 2 to 3, and 15 healthy controls were exposed to multidirectional translational stance perturbations on a moveable platform. Perturbations were unpredictable in terms of amplitude, timing and direction. Patients were tested in the medication ON and OFF (at least 12 h of dopaminergic medication withdrawal) state on two separate days. Forward and backward stepping responses were quantified in terms of (1) presence, onset and amplitude of anticipatory postural adjustments (APAs); (2) spatiotemporal step variables (step onset, length and velocity); and (3) leg inclination angle at first stepping-foot contact. When perturbed forward, patients performed worse than controls in terms of step length (0.32 ± 0.07 vs. 0.38 ± 0.05 m, p = 0.01) and step velocity (1.21 ± 0.16 vs. 1.37 ± 0.13 m/s, p = 0.01), while step onset was not different. The number of steps with an APA was larger in patients in the OFF state than in controls which was, however, only significant after forward perturbations (43 vs. 20%, p = 0.01). Following backward perturbations, leg angles at foot contact were smaller in patients compared to controls (-2.71° ± 4.29° vs. 0.26° ± 2.80°, p = 0.04) reflecting a poorer mechanical efficiency of the step. Dopaminergic medication had no significant effect on any of these outcomes. In conclusion, dopaminergic medication does not improve underscaling of stepping responses in PD. Therefore, other interventions are needed to improve these important defense postural reactions.