OBJECTIVE: To determine safety of intramuscular botulinum toxin A (BoNT-A) injections to reduce spasticity and improve care and comfort of nonambulatory children with cerebral palsy (CP). METHODS: Nonambulatory children with CP were randomly allocated to receive either BoNT-A (n = 23) or sham procedure (n = 18) in Cycle 1. In Cycle 2, the BoNT-A group received a second episode of BoNT-A (n = 20) and sham group received their first episode of BoNT-A (n = 17). A pediatric rehabilitation specialist masked to group allocation graded each adverse event (AE) according to system, severity (mild, moderate, serious, sentinel) and causality (unlikely/unrelated; possible; probable/definite). RESULTS: There was no difference for all moderate/serious AEs between the BoNT-A and sham/control groups in either Cycle 1 (incident rate ratio = 1.30, 95% confidence interval = 0.43–4.00; P = .64) or Cycle 2 (incident rate ratio = 0.72, 95% confidence interval = 0.30–1.75; P = .47). In Cycle 2, 1 serious, 3 moderate (single-episode group), and 24 mild (single-episode group n = 10; 2 episode group n = 14) AEs were probably/definitely related to BoNT-A. CONCLUSIONS: Children receiving BoNT-A were at no greater risk of moderate/serious AEs compared with a sham control procedure. There was no increased risk of moderate/serious AEs between one and two episodes of BoNT-A.