Botulinum toxin (BT) is an effective treatment for spasticity symptoms in multiple sclerosis (MS). Despite its wide use in clinical practices, only few studies have explored long-term persistence. We aim to evaluate the rate of discontinuation of BT treatment and the correlation with MS, spasticity, and injection variables. This retrospective study on 3-year prospectively collected data included 122 MS patients receiving BT injections for spasticity. We collected MS clinical variables (disease durations, Expanded Disability Status Scales [EDSSs], disease-modifying treatments [DMT], and Symbol Digit Modalities Tests [SDMTs]), modified Ashworth scales [MASs], concomitant treatments, and injection variables (formulation, dose, number of injections, and intervals between injections). A total of 14 out of the 122 patients discontinued BT after a mean time of 3.0 ± 1.5 years. In the Cox regression model including the MS clinical variables, the probability of BT discontinuations increased in patients with DMT changes during follow-ups (HR = 6.34; 95%Cl = 2.47, 18.08; p < 0.01) and with impaired SDMTs (HR = 1.20; 95%Cl = 1.04, 1.96; p < 0.01). In the model including the spasticity variables, there were no associations between BT discontinuation and MAS or other spasticity treatments. In the model including the injection variables, the probability of discontinuation decreased by 80% for each cumulative injection (HR = 0.16; 95%Cl = 0.05, 0.45; p < 0.01), but increased by 1% for each additional day over the 3-month interval between injections (HR = 1.27; 95%Cl = 1.07, 1.83; p < 0.01). BT discontinuation was associated with concomitant MS-related issues (e.g., treatment failure and DMT change) and the presence of cognitive impairment, which should be accounted for when planning injections. The interval between injections should be kept as short as possible from regulatory and clinical perspectives to maximize the response across all of the spasticity symptoms and to reduce discontinuation in the long term.