SUMMARY Saliva carbamazepine and phenytoin samples were used to monitor treatment in 35 children aged between 2 and 14 years during a 2-year period. All phenytoin levels and over half the carbamazepine levels that were above the therapeutic range were associated with adverse effects. Dose and carbamazepine saliva levels were significantly related but no such relationship was found for phenytoin. There was no apparent relationship between the saliva level of either drug and convulsion control. (81 %), and between 0800 and 1600 hours, 115 (54%) samples were collected.Saliva was assayed for CBZ by a modification4 of the gas-liquid chromatography method for serum described by Least et al.5 and for DPH by radioimmunoassay6 using antisera supplied by Dr G W Aherne, University of Surrey.The number of samples taken from each child varied. In those children who gave at least 5 samples, intra-individual variation on different dose frequency regimens was compared using Wilcoxon's rank sum test for unpaired samples.CBZ and DPH level-dose relationships were determined using all samples and also using those samples collected at 3 and 5 hours after dose administration.Results