We present a patient with severe pyridox(am)ine 5'-phosphate oxidase deficiency and homozygosity for a novel nonsense-mutation, p.A174X, in the PNPO gene who died with pyridoxal phosphate (PLP) treatment despite initial clinical recovery. He presented neonatally, with the classical clinical symptoms of the disease. Increase of urinary vanillactate was the first biochemical factor of alert. Amino acid and neurotransmitter analysis in CSF indicated reduced activity of several PLP-dependent enzymes. The diagnosis was confirmed by mutational studies. From this and the other reported patients it may be concluded that the administration of PLP should not be delayed until the complete biochemical evidence is obtained.