Aim  The long‐term outcome of the Dutch pyridoxine‐dependent epilepsy cohort and correlations between patient characteristics and follow‐up data were retrospectively studied.Method  Fourteen patients recruited from a national reference laboratory were included (four males, 10 females, from 11 families; median age at assessment 6y; range 2y 6mo–16y). The following data were retrieved: sex; age at seizure onset; age at the start of pyridoxine therapy; level of urinary alpha‐aminoadipic semialdehyde; antiquitin mutations; developmental milestones; evaluation of neurocognitive functioning and school career; magnetic resonance imaging (MRI) and electroencephalography (EEG) assessments.Results  Pyridoxine was started antenatally in two children, in the first week of life in five, in the first month of life in three, or after the first month of life (range 2.5–8mo) in four. No child was physically disabled; however, only five walked at 2 years of age. Mental development was delayed in most: median IQ or developmental index was 72 (SD 19). Pyridoxine monotherapy controlled seizures in 10 of 14 children, whereas four needed additional antiepileptic drugs. Seizure persistence, antiepileptic drugs (other than pyridoxine), EEG background, and epileptiform activity were not associated with outcome. On neonatal MRI, structural and white matter abnormalities occurred in five of eight children; on follow‐up, the number of abnormal MRIs was increased. Delayed initiation of pyridoxine medication and corpus callosum abnormalities were significantly associated with unfavourable neurodevelopmental outcome, but normal follow‐up imaging did not predict a good outcome.Interpretation  Outcome of patients with pyridoxine‐dependent epilepsy remains poor. Individual outcome cannot be predicted by the evaluated characteristics. We suggest that collaborated research in structured settings could help to improve treatment strategies and outcome for pyridoxine‐dependent epilepsy.